Aalto University Theory of Superconductivity

Theory of Superconductivity
Tfy-3.4801 (6 cr.) P, III.

Aalto University, School of Science and
Technology
Department of Applied Physics
Fall, 2010
N. B. Kopnin
Low Temperature Laboratory, Aalto University
email: kopnin@boojum.hut.
August 26, 2010
2
Contents
1 Introduction to Superconductivity 7
1.1 Superconducting transition . . . . . . . . . . . . . . . . . . . . . 7
1.2 The London model . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.1 Meissner eect . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Phase coherence . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.1 Magnetic ux quantization . . . . . . . . . . . . . . . . . 11
1.3.2 Coherence length and the energy gap . . . . . . . . . . . . 12
1.4 Critical currents and magnetic elds . . . . . . . . . . . . . . . . 13
1.4.1 Condensation energy . . . . . . . . . . . . . . . . . . . . . 13
1.4.2 Critical currents . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 Quantized vortices . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5.1 Basic concepts . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5.2 Vortices in the London model . . . . . . . . . . . . . . . . 16
1.5.3 Critical elds in typeII superconductors . . . . . . . . . 17
1.5.4 The Little and Parks eect . . . . . . . . . . . . . . . . . 19
2 The BCS theory 23
2.1 Landau Fermi-liquid . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 The Cooper problem . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 The BCS model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.3.1 Wave functions . . . . . . . . . . . . . . . . . . . . . . . . 31
2.3.2 Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.3 The Bogoliubovde Gennes equations . . . . . . . . . . . 34
2.3.4 The self-consistency equation . . . . . . . . . . . . . . . . 36
2.4 Observables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.4.1 Energy spectrum and coherence factors . . . . . . . . . . 39
2.4.2 The energy gap . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4.3 Condensation energy . . . . . . . . . . . . . . . . . . . . . 43
2.4.4 Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.4.5 Negative energies . . . . . . . . . . . . . . . . . . . . . . . 48
2.5 Impurities. Anderson theorem . . . . . . . . . . . . . . . . . . . . 48
3
4 CONTENTS
3 Andreev reection 53
3.1 Semiclassical approximation . . . . . . . . . . . . . . . . . . . . . 53
3.1.1 Andreev equations . . . . . . . . . . . . . . . . . . . . . . 53
3.1.2 Andreev reection . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Andreev states . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2.1 SNS structures . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3 Supercurrent through an SNS structure. Proximity eect . . . . 63
3.3.1 Short junctions. Point contacts . . . . . . . . . . . . . . . 63
3.3.2 Long junctions . . . . . . . . . . . . . . . . . . . . . . . . 66
4 SuperconductorInsulatorNormal-metal Interface 73
4.1 Transmission through the barrier . . . . . . . . . . . . . . . . . . 73
4.1.1 Transmission and reection probabilities . . . . . . . . . . 73
4.1.2 Probability conservation . . . . . . . . . . . . . . . . . . . 79
4.2 Current through the NIS junction . . . . . . . . . . . . . . . . . . 80
4.2.1 Normal tunnel resistance . . . . . . . . . . . . . . . . . . 83
4.2.2 Landauer formula . . . . . . . . . . . . . . . . . . . . . . 84
4.2.3 Tunnel current . . . . . . . . . . . . . . . . . . . . . . . . 85
4.2.4 Excess current . . . . . . . . . . . . . . . . . . . . . . . . 86
4.2.5 NS Andreev current. Current conversion . . . . . . . . . . 87
4.3 SIS contact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.3.1 Wave functions and the energy of bound states . . . . . . 87
4.3.2 Supercurrent . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.4 Scattering matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.4.1 SINIS structures . . . . . . . . . . . . . . . . . . . . . . . 91
4.4.2 Interference eects in short contacts . . . . . . . . . . . . 93
5 Weak links 95
5.1 Josephson eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.1.1 D.C and A.C. Josephson eects . . . . . . . . . . . . . . . 95
5.1.2 Superconducting Quantum Interference Devices . . . . . . 98
5.2 Josephson vortices in extended junctions. . . . . . . . . . . . . . 98
5.2.1 Low eld limit . . . . . . . . . . . . . . . . . . . . . . . . 100
5.2.2 Higher elds. Josephson vortices. . . . . . . . . . . . . . . 101
5.3 Dynamics of Josephson junctions . . . . . . . . . . . . . . . . . . 103
5.3.1 Resistively shunted Josephson junction . . . . . . . . . . . 103
5.3.2 The role of capacitance . . . . . . . . . . . . . . . . . . . 104
5.3.3 Thermal uctuations . . . . . . . . . . . . . . . . . . . . . 109
5.3.4 Shapiro steps . . . . . . . . . . . . . . . . . . . . . . . . . 111
6 Coulomb blockade in normal double junctions 115
6.1 Orthodox description of the Coulomb blockade . . . . . . . . . . 115
6.1.1 Low temperature limit . . . . . . . . . . . . . . . . . . . . 118
6.1.2 Conductance in the high temperature limit . . . . . . . . 119
CONTENTS 5
7 Quantum phenomena in Josephson junctions 123
7.1 Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7.1.1 Quantum conditions . . . . . . . . . . . . . . . . . . . . . 123
7.1.2 Charge operator . . . . . . . . . . . . . . . . . . . . . . . 124
7.1.3 The Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . 126
7.2 Macroscopic quantum tunnelling . . . . . . . . . . . . . . . . . . 127
7.2.1 Eects of dissipation on MQT . . . . . . . . . . . . . . . 129
7.3 Band structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.3.1 Blochs theorem . . . . . . . . . . . . . . . . . . . . . . . 130
7.3.2 Blochs theorem in Josephson devices . . . . . . . . . . . 131
7.3.3 Large Coulomb energy: Free-phase limit . . . . . . . . . . 132
7.3.4 Low Coulomb energy: Tight binding limit . . . . . . . . . 134
7.4 Coulomb blockade . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.4.1 Equation of motion . . . . . . . . . . . . . . . . . . . . . . 136
7.4.2 Bloch oscillations and the Coulomb blockade in Josephson
junctions . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.4.3 Eect of dissipation . . . . . . . . . . . . . . . . . . . . . 138
7.5 Parity eects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6 CONTENTS
Chapter 1
Introduction to
Superconductivity
1.1 Superconducting transition
Superconductivity manifests itself mainly as an absence of resistivity below some
critical temperature. It was discovered in 1911 by H. Kamerlingh Onnes in
Leiden, three years after he rst liqueed
4
He. He measured the resistivity of
mercury. The resistivity behavior as a function of temperature is shown in Fig.
1.1.
T T
c
n
= 0
Figure 1.1: Below the transition temperature, the resistivity drops to zero.
The absolute absence of resistivity is a very fundamental phenomenon. In
combination with general quantum-mechanical principles, it can lead to quite
informative conclusions on the properties of the superconducting state. Here we
try to describe the basic picture of superconductivity using minimum amount of
input information. We consider the most striking properties of superconductors
such as their ideal diamagnetism, macroscopic quantum nature of superconduc-
tivity including phase coherence which leads to zero resistivity, to quantization
of magnetic ux and to formation of quantized vortices. The maximum values
of magnetic elds and currents that can be withstood by superconductors are
7
8 CHAPTER 1. INTRODUCTION TO SUPERCONDUCTIVITY
Table 1.1: Parameters for metallic superconductors
T
c
, K H
c
, Oe H
c2
, Oe
L
,

A
0
,

A Type
Al 1.18 105 500 16000 0.01 I
Hg 4.15 400 400 I
Nb 9.25 1600 2700 470 390 1.2 II
Pb 7.2 800 390 830 0.47 I
Sn 3.7 305 510 2300 0.15 I
In 3.4 300 400 3000 I
V 5.3 1020 400 300 0.7 II
Table 1.2: Parameters for some high temperature superconductors
T
c
, K H
c2
, T
L
,

A
0
,

A Type
Nb
3
Sn 18 25 2000 115 II
La
0.925
Sr
0.072
CuO
4
34 1500 20 75 II
YBa
2
Cu
3
O
7
92.4 150 2000 15 140 II
Bi
2
Sr
2
Ca
3
CuO
10
111 II
Tl
2
Sr
2
Ca
2
Cu
3
O
10
123 II
HgBa
2
Ca
2
Cu
3
O
8
133 II
MgB
2
36.7 14 1850 50 40 II
also briey discussed. The rest of the course is devoted to a microscopic theory
of superconductivity.
1.2 The London model
We assume that the current ows without dissipation and has the form
j
s
= n
s
ev
s
whence the velocity of superconducting electrons is v
s
= j/n
s
e where n
s
is
their density. Now we come to the most important argument [F. London and
H. London, 1935]: Being non-dissipative, this current contributes to the kinetic
energy of superconducting electrons. The total free energy is a sum of the kinetic
energy of superconducting electrons and the magnetic energy
T =
_ _
n
s
mv
2
s
2
+
h
2
8
_
dV =
_ _
mj
2
s
2n
s
e
2
+
h
2
8
_
dV .
Here h is the microscopic magnetic eld. Its average over large area in the
sample gives the magnetic induction B. Using the Maxwell equation
j
s
= (c/4)curl h , (1.1)
1.2. THE LONDON MODEL 9
we transform this to the following form
T =
_ _
mc
2
32
2
n
s
e
2
(curl h)
2
+
h
2
8
_
dV =
1
8
_
dV
_
h
2
+
2
L
(curl h)
2
_
. (1.2)
Here
L
=
_
mc
2
4n
s
e
2
_
1
2
(1.3)
is called the London penetration depth. In equilibrium, the free energy is min-
imal with respect to distribution of the magnetic eld. Variation with respect
to h gives
T =
1
4
_
dV
_
h h +
2
L
(h) (h)
=
1
4
_
dV
_
h +
2
L
h
_
h +
1
4
_
dV div [h curl h] .
Here we use the identity
div [b a] = a [b] b [a]
and put a = h, b = h. Looking for a free energy minimum and omitting
the surface term we obtain the London equation:
h +
2
L
curl curl h = 0 . (1.4)
Since
curl curl h = div h
2
h
and div h = 0, we nd
h
2
L
2
h = 0 . (1.5)
1.2.1 Meissner eect
Equation (1.5) in particular describes the Meissner eect, i.e., an exponential
decay of weak magnetic elds and supercurrents in a superconductor. The
characteristic length over which the magnetic eld decreases is just
L
. Consider
a superconductor which occupies the half-space x > 0. A magnetic eld h
y
is
applied parallel to its surface (Fig. 1.2). We obtain from Eq. (1.5)
2
h
y
x
2

2
L
h
y
= 0
which gives
h
y
= h
y
(0) exp(x/) .
The eld decays in a superconductor such that there is no eld in the bulk.
According to Eq. (1.1) the supercurrent also decays and vanishes in the bulk.
h
h
y
x
0
S
L
Figure 1.2: The Meissner eect: Magnetic eld penetrates into a superconductor
only over distances shorter than
L
.
Therefore,
B = H+ 4M = 0
in a bulk superconductor, where H is the applied led. The magnetization and
susceptibility are
M =
H
4
; =
M
H
=
1
4
(1.6)
as for an ideal diamagnetic: Superconductor repels magnetic eld lines. The
Meissner eect in type I superconductors persists up to the eld H = H
c
(see
Table 1.1, Fig. 1.7, and the section below) above which superconductivity is
destroyed. Type II superconductors display the Meissner eect up to much
lower elds, after which vortices appear (see Section 1.5).
1.3 Phase coherence
The particle mass ow is determined by the usual quantum-mechanical expres-
sion for the momentum per unit volume
j
m
=
ih
2
[
] = h[[
2
. (1.7)
In order to have a nite current in the superconductor it is necessary that
is the wave function of all the superconducting electrons with a denite phase :
the superconducting electrons should all be in a single quantum state. According
to the present understanding what happens is that the electrons (Fermi parti-
cles) combine into pairs (Cooper pairs, see the next Chapter) which are Bose
objects and condense into a Bose condensate. The current appears when the
phase of the condensate function slowly varies in space. Equation (1.7) sug-
gests that P = h is the momentum of a condensate particle (which is a pair in
the superconductor). For charged particles, the momentum is p = P(e
/c)A
where P is the canonical momentum, A is the vector potential of the magnetic
1.3. PHASE COHERENCE 11
B
l
Figure 1.3: Magnetic ux through the hole in a superconductor is quantized.
eld, and e
is the charge of the carrier. In superconductors the charge is carried

by pairs of electrons thus e
= 2e and the Cooper pair mass is 2m.

Using the denition of the momentum we introduce the velocity of super-
conducting electrons
v
s
=
h
2m
_

2e
hc
A
_
. (1.8)
Now the electric current becomes
j
s
= n
s
ev
s
=
e
2
n
s
mc
_
A
hc
2e
_
. (1.9)
where [[
2
= n
s
/2 is the density of electron pairs.
It is instructive to compare this equation with Eqs. (1.1) and (1.4). We nd
from these
curl j =
c
4
curl curl h =
c
4
2
L
h =
c
4
2
L
curl A
Therefore,
j =
c
4
2
L
(A) =
e
2
n
s
mc
(A)
where is a gradient of some function. It is seen that this coincides with Eq.
(1.9) where = ( hc/2e).
1.3.1 Magnetic ux quantization
Let us consider an non-singly-connected superconductor with dimensions larger
than
L
placed in a magnetic eld (Fig. 1.3). We choose a contour which goes
all the way inside the superconductor around the hole and calculate the contour
integral
_ _
A
hc
2e
_
dl =
_
S
curl A dS
hc
2e
=
hc
2e
2n . (1.10)
Here is the magnetic ux through the contour. The phase change along the
closed contour is = 2n where n is an integer because the wave function
is single valued. Since j = 0 in the bulk, the l.h.s. of Eq. (1.10) vanishes, and
we obtain =
0
n where
0
=
hc
e
2.07 10
7
Oe cm
2
(1.11)
is the quantum of magnetic ux. In SI units,
0
= h/e = 2.07 10
15
Tm
2
.
1.3.2 Coherence length and the energy gap
Cooper pairs keep their correlation within a certain distance called the coher-
ence length (see the next Chapter). This length introduces an important
energy scale. To see this let us argue as follows. Since the correlation of pairs
is restricted within the phase gradient cannot exceed 1/; thus the super-
conducting velocity cannot be larger than the critical value
v
c
=
h
m
. (1.12)
where 1 is a constant. Thus the energy of a correlated motion of a pair is
restricted to
0
p
F
v
c
= hv
F
/. This gives

hv
F
0
.
The quantity
0
is in fact the value of the energy gap (0) at zero temper-
ature in the single-particle excitation spectrum in the superconducting state.
We shall see from the microscopic theory in the next Chapter that the energy
of excitations
=
_
p
2
2m
E
F
_
2
+
2
cannot be smaller than a certain value that generally depends on temperature.
The coherence length is usually dened as
0
=
hv
F
2k
B
T
c
where
0
= 1.76k
B
T
c
and
0
is the coherence length at zero temperature of a
clean (without impurities) material. In alloys with <
0
,
=
_
where is the mean free path.

The ratio
=

L
1.4. CRITICAL CURRENTS AND MAGNETIC FIELDS 13

is called the GinzburgLandau parameter. Its magnitude separates all super-
conductors between type-I ( < 1/
2 0.7) and type-II ( > 1/
2) supercon-
ductors. For alloys with <
0
= 0.75
where
0
is the London length in a clean material at zero temperature. The
conclusion is that alloys are type-II superconductors. Values of
L
,
0
, and
for some materials are listed in Tables 1.1 and 1.2.
1.4 Critical currents and magnetic elds
1.4.1 Condensation energy
The kinetic energy density of condensate (superconducting) electrons cannot
exceed
F
c
=
n
s
mv
2
c
2
=
n
s
h
2
2
2
m
2
. (1.13)
If the velocity v
s
increases further, the kinetic energy exceeds the energy gain
of the superconducting state with respect to the normal state F
n
F
s
, and
superconductivity disappears. Therefore, F
c
= F
n
F
s
is just this energy gain
which is called the condensation energy.
Assume now that the superconductor is placed in a magnetic eld H. It
repels the eld thus increasing the energy of the external source that creates
the eld. The energy of the entire system increases and becomes
F = F
s
+
H
2
8
= F
n
F
c
+
H
2
8
.
In the superconducting state, F < F
n
. When the energy reaches the energy of
a normal state F
n
, the superconductivity becomes no longer favorable energet-
ically. Thus the thermodynamic critical magnetic eld satises
F
c
=
H
2
c
8
.
Using the expression for
L
we nd from Eq.(1.13)
H
c
=
hc
e
L
=

0
The exact expression for H

c
at temperatures close to T
c
is
H
c
=

0
2
2
L
(1.14)
Values of H
c
for some materials are given in Table 1.1.
1.4.2 Critical currents
There may be several mechanisms of destruction of superconductivity by a cur-
rent owing through it.
Mechanism 1. Large type-I samples: The critical current I
c
creates H
c
at
the sample surface. For a cylinder with a radius R,
2RH
c
=
4
c
I
c
.
If R
L
, the current ows only within the layer of a thickness
L
near the
sample surface. Thus I
c
= 2R
L
j
c
and
j
c
=
cH
c
4
L
. (1.15)
Mechanism 2. If the transverse dimensions of the superconductor a and b
are small, a, b
L
the current is distributed uniformly over the cross section
of the sample. In this case, the dominating mechanism is the pair-breaking:
superconductivity is destroyed by the high velocity of superconducting electrons.
The critical current is
j
c
= n
s
ev
c
= n
s
eh/m . (1.16)
In fact, this current density coincides with the critical current in thick samples.
Indeed, inserting H
c
and
L
in Eq. (1.15) we obtain Eq. (1.16). However, the
magnetic eld created at the surface H (c/4)j
c
a
2
/a (a/
L
)H
c
is smaller
than H
c
. The critical current in Eqs. (1.15), (1.16) is very high. For H
c
= 500
Oe and
L
= 500

A it can be as high as 10
8
A/cm
2
.
In type-II superconductors, critical magnetic elds and currents are associ-
ated with quantized vortices.
1.5 Quantized vortices
1.5.1 Basic concepts
Consider a type-II superconductor where the London length is large. The super-
current and magnetic eld do not vanish within the region of the order of from
the surface: there exists a sizable region of nonzero v
s
. If the magnetic eld is
large enough, v
s
can reach high values, v
s
(e/mc)Hr. For elds H hc/e
2
,
the velocity can reach rh/m
2
v
c
for r . This would lead to destruction of
superconductivity if there were no means for compensating a large contribution
to v
s
due to the magnetic eld.
Assume that we have a linear singularity such that the phase of the wave
function of superconducting electrons changes by 2n if one goes around this
lines along a closed contour, see Fig. 1.4. Consider again the integral along this
contour
mc
e
_
v
s
dl =
_ _
A
hc
2e
_
dl =
_
S
curl A dS
hc
2e
1.5. QUANTIZED VORTICES 15

B
Figure 1.4: Singular lines in a SC with 2n phase variations around them.
=
hc
2e
2n =
0
n . (1.17)
Here is the magnetic ux through the contour,
0
is the ux quantum. The
phase change along the closed contour is = 2n. We observe that the super-
conducting velocity increase is completely compensated by the phase variation
if the magnetic ux is =
0
n. One can thus expect that in superconductors
with large
L
in high magnetic elds, there will appear linear singularities with
a surface density n
L
such that the magnetic induction
B = (
0
n)n
L
(1.18)
is all distributed over vortices. Under these conditions, the superconducting
velocity does not increase with distance, and superconductivity is conserved on
average.
Each singularity of the phase can exist if the wave function of the supercon-
ducting electrons, i.e., the density of superconducting electrons n
s
= 2[[
2
goes
to zero at the singular line. The size of the region where n
s
is decreased with
respect to its equilibrium value has a size of the order of the coherence length
and is called the vortex core. Such singular objects are called quantized vortices:
each vortex carries a quantized magnetic ux
0
n. The condition required for
existence of vortices is > or exactly > 1/
2. More favorable energetically

are singly quantized vortices which carry one magnetic ux quantum and have
a phase circulation 2 around the vortex axis.
Vortices are the objects which play a very special role in superconductors and
superuids. In superconductors, each vortex carries exactly one magnetic-ux
quantum. Being magnetically active, vortices determine the magnetic properties
of superconductors. In addition, they are mobile if the material is homogeneous.
In fact, a superconductor in the vortex state is no longer superconducting in a
usual sense. Indeed, there is no complete Meissner eect: some magnetic eld
penetrates into the superconductor via vortices. In addition, regions with the
normal phase appear: since the order parameter turns to zero at the vortex axis
and is suppressed around each vortex axis within a vortex core with a radius
of the order of the coherence length, there are regions with a nite low-energy
density of states. Moreover, mobile vortices come into motion in the presence
of an average (transport) current. This produces dissipation and causes a nite
resistivity (the so-called ux ow resistivity): a superconductor is no longer
superconducting.
To avoid motion of vortices and thus ensure zero resistance of a superconduc-
tor, various defects such as granular structure, lattice defects, articial defects
are introduced into the superconducting material. These defects attract vor-
tices, or pin them in the superconductor. To overcome the pinning force one
has to apply a nite current density, critical depinning current j
c
, that produces
the Lorentz force
F
L
=

0
c
[j
c
z]
where z is the unit vector in the direction of the magnetic eld. Depending
on the material, the critical current can be as high as 10
4
10
5
A/cm
2
. For
currents below the depinning current, a type-II superconductor can have zero
resistance up to very high magnetic elds H
c2
which are considerably higher
than H
c
(see below).
In superuids, vortices appear in a container with helium rotating at an
angular velocity above a critical value which is practically not high and can
easily be reached in experiment. Vortices are also created if a superuid ows in
a tube with a suciently high velocity. The driving force that pushes vortices is
now the Magnus force. Vortices move and experience reaction from the normal
component; this couples the superuid and normal components and produces
a mutual friction between them. As a result, the superow is no longer
persistent.
1.5.2 Vortices in the London model
Let us take curl of Eq. (1.9). We nd
h
hc
2e
curl =
mc
n
s
e
2
curl j
s
=
2
L
curl curl h .
This looks like Eq. (1.5) except for one extra term. This term is nonzero if
there are vortices. In the presence of vortices, the London equation should be
modied. For an n-quantum vortex we have
curl = 2n z
(2)
(r)
where z is the unit vector in the direction of the vortex axis. Therefore, the
London equation for a vortex becomes
h +
2
L
curl curl h = n
0
(2)
(r) (1.19)
where
0
is the vector along the vortex axis with the magnitude of one ux
quantum. For a system of vortices
h +
2
L
curl curl h = n
0
(2)
(r r
k
) (1.20)
h
r
L
Figure 1.5: Structure of a single vortex. The core region with the radius
is surrounded by currents. Together with the magnetic eld, they decay at
distances of the order of
L
.
where the sum is over all the vortex positions r
k
.
One can easily nd the magnetic eld for a single straight vortex (see Prob-
lem 1.1). In cylindrical coordinates h = (0, 0, h
z
(r)), the magnetic eld is
h
z
(r) =
n
0
2
2
L
ln
_
L
r
_
near the vortex axis r
L
. The magnetic eld increases logarithmically near
the vortex axis. However, in our model, the coordinate r cannot be made shorter
than the coherence length since n
s
vanishes at the vortex axis, and the London
equation does not apply for r < . Therefore, at the axis
h(0) =
n
0
2
2
L
ln
_
_
.
We can calculate the current around the vortex near the core.
j
=
c
4
h
z
r
=
nc
0
8
2
2
L
r
=
n
s
enh
2mr
For a single-quantum vortex the superconducting velocity is
v
s,
=
h
2mr
Therefore, the phase is just the azimuthal angle:
=
1.5.3 Critical elds in typeII superconductors
The free energy of a single-quantum vortex per unit length is
T =
1
8
_
_
h
_
h +
2
L
curl curl h
_
+ div[h curl h]
d
2
r
=
1
8
_
h
z
(2)
(r) d
2
r +
1
8
_
[h curl h] dl . (1.21)
The last integral is taken along a remote contour and vanishes. The rst integral
gives
T
L
=

2
0
16
2
2
L
ln

L
=
h
2
n
s
4m
ln

L
. (1.22)
For an n-quantum vortex we would obtain
T
L
=
n
2
2
0
16
2
2
L
ln
_
_
. (1.23)
The energy is proportional to n
2
. Therefore, vortices with n > 1 are not favor-
able energetically. For given induction B, the number of n-quantum vortices
is proportional to n
1
, while the energy of each vortex is proportional to n
2
.
Therefore, the energy of the vortex array grows as n for given induction B.
Therefore, the vortex array of single-quantum vortices has lower energy than
the array of n-quantum vortices.
Equation (1.22) allows to nd the lower critical magnetic eld, i.e., the eld
H above which the rst vortex appears. The free energy of a unit volume
of a superconductor with a set of single-quantum vortices is F
L
= n
L
T
L
=
(B/
0
)T
L
. The proper thermodynamic potential in an external eld H is the
Gibbs free energy G = F HB/4
G =
BT
L
BH
4
=
B
0
16
2
2
L
ln
_
BH
4
.
If H < H
c1
where
H
c1
=

0
4
2
L
ln
_
_
, (1.24)
the Gibbs free energy is minimal for B = 0. This corresponds to zero eld in
the bulk: the Meissner eect takes place. The free energy becomes negative if
H > H
c1
. Therefore, it decreases with increasing B inside the superconductor.
This means that vortices appear for H > H
c1
.
As the magnetic eld increases, vortices become more and more dense, see
Eq. (1.18), and the normal phase in the cores occupies larger and larger fraction
of the sample. The superconductivity is totally destroyed when their cores start
to overlap, i.e., when their density n
L
= B/
0
1/
2
. The exact condition is
H
c2
=

0
2
2
.
Using Eq. (1.14) we note that
H
c1
= H
c
ln
2
; H
c2
=
2H
c
, (1.25)
i.e., for superconductors with a large , the critical eld H
c1
is considerably
lower than H
c2
. At the same time, the upper critical eld H
c2
is considerably
higher than H
c
.
The phase diagram of a type II superconductor is shown in Fig. 1.6.
For more reading on vortices in type II superconductors see Refs. [5, 6, 7, 8].
Meissner
Vortex
Normal
H
T
T
H
H
H
c
c
c1
c2
Figure 1.6: Phase diagram of a type II superconductor
M
H H
c1
c2
H
H
c
Figure 1.7: Full line: Magnetization of a type II superconductor. The linear
part at low elds corresponds to the full Meissner eect Eq. (1.6). Dashed line:
Magnetization of a type I superconductor. The Meissner eect persists up to
the thermodynamic critical eld H
c
.
1.5.4 The Little and Parks eect
Let us consider a thin lm with thickness d deposited on an insulating cylinder
of radius R and length L R. It is placed into an external magnetic eld H
parallel to the cylinder axis, see Fig. 1.8. We assume that d
L
and d R.
After the superuid transition a current in an azimuthal direction is generated
in the lm. The current is distributed homogeneously over the thickness of the
lm since d
L
. The magnetic led inside the cylinder is homogeneous, as
well.
Free energy of the system per unit length of the cylinder is
T =
_
d
2
r
_
H
2
in
8
+
mj
2
s
2n
s
e
2
_
=
H
2
in
R
2
8
+
mj
2
s
2Rd
2n
s
e
2
L
2R
H
d
H
in
Figure 1.8: Superconducting lm of thickness d is deposited on an insulating
cylinder of radius R and placed into magnetic eld H.
Using the solution of problem 1.6 we have
H
in
_
1 +
2
2
dR
_
=
n
0
R
2
+
2
2
dR
H (1.26)
and
j
s
=
c
4d(1 + 2
2
/dR)
_
H
n
0
R
2
_
Therefore, after a little algebra the free energy becomes
T =
R
2
8(1 + 2
2
/dR)
_
_
n
0
R
2
_
2
+
2
2
H
2
dR
_
As we know, in an applied magnetic led H, it is the Gibbs free energy (per
unit length)
G = T
_
d
2
r
HB
4
=
R
2
8(1 + 2
2
/dR)
_
_
n
0
R
2
_
2
+
2
2
H
2
dR
_
HH
in
R
2
4
which has to be minimized. Using Eq. (1.26) we nd
G =
R
2
8(1 + 2
2
/dR)
_
n
0
R
2
H
_
2
H
2
R
2
8
The last term here is the Gibbs free energy of the cylinder interior without the
superconductor. It can thus be ignored.
The rest term is the free energy G of the superconductor. It is minimal
when the applied magnetic eld introduces an integer number of ux quanta
G
R
2
H/
0
1 2 3 0
Figure 1.9: Gibbs free energy of a superconducting lm deposited on an insu-
lating cylinder and placed into magnetic eld H. The plot is a set of parabolas
shifted by integer numbers of ux quanta along the horizontal axis. The free en-
ergy oscillates as it goes from one parabola to another with increasing magnetic
eld.
into the cylinder. The free energy as a function of the magnetic eld is shown
in Fig. 1.9. If the ux quanta can easily pass through the superconducting
lm to or from the interior of the cylinder, the system will choose the branch
with the corresponding number of ux quanta which minimizes the free energy
for given H. The free energy thus oscillates as a function of the magnetic eld
due to entrance of ux quanta into the cylinder. The critical temperature of
the superconductor oscillates in a similar manner. This is called the Little and
Parks eect.
Problems
Problem 1.1
Calculate the magnetic eld and current around a single straight vortex.
Problem 1.2
Calculate the energy of a circular vortex loop with a radius R such that
R
L
.
Problem 1.3
Find the distribution of the magnetic eld and of the current in a supercon-
ducting slab of a thickness d placed in an external homogeneous magnetic eld
H parallel to the slab. The longitudinal dimensions of the slab are much larger
than d.
Problem 1.4
Find the distribution of the magnetic eld and of the current in a supercon-
ducting slab of a thickness d carrying a total current I along the length of the
slab. The longitudinal dimensions of the slab are much larger than d.
Problem 1.5
A thin superconducting lm with a thickness d
L
is deposited on a
dielectric lament (cylinder) of a radius R d. The lament is placed into
a longitudinal magnetic eld at a temperature T > T
c
and then cooled down
below T
c
. The eld is then switched o. Find out how the captured magnetic
eld is quantized.
Problem 1.6
The same setting as in Problem 1.5: A thin superconducting lm with a
thickness d
L
is deposited on a dielectric lament (cylinder) of a radius R
d. The lament is placed into a longitudinal magnetic eld H at a temperature
T > T
c
and then cooled down below T
c
. Calculate the captured magnetic eld
inside the dielectric cylinder and the supercurrent in the lm.
Chapter 2
The BCS theory
2.1 Landau Fermi-liquid
The ground state of a system of Fermions corresponds to the lled states with
energies E below the maximal Fermi energy E
F
, determined by the number
of Fermions. In an homogeneous system, one can describe particle states by
momentum p such that the spectrum becomes E
p
. The condition of maximum
energy E
p
= E
F
denes the Fermi surface in the momentum space. In an
isotropic system, this is a sphere such that its volume divided by (2h)
3
n
=
4p
3
F
3(2h)
3
gives number of particles with the spin projection per unit (spatial) volume
of system. For electrons with spin
1
2
, the total number of particles in the unit
volume of the system, i.e., the particle density is twice n
n =
p
3
F
3
2
h
3
(2.1)
This ground state corresponds to a ground-state energy E
0
.
Excitations in the Fermi liquid that increase its energy as compared to E
0
are created by moving a particle from a state below the Fermi surface to a state
above it. This process can be considered as a superposition of two processes.
First is a removal of a particle from the system out of a state below the Fermi
surface. The second is adding a particle to a state above the Fermi surface. By
taking a particle out of the state with an energy E
1
< E
F
we increase the energy
of the system and create a hole excitation with a positive energy
1
= E
F
E
1
.
By adding a particle into a state with an energy E
2
> E
F
we again increase the
energy and create a particle excitation with a positive energy
2
= E
2
E
F
.
The energy of the system is thus increased by
1
+
2
= E
2
E
1
.
Shown in Fig. 2.1 are processes of creation of particle and hole excitations
in a Fermi liquid. Consider it in more detail. Removing a particle with a
23
24 CHAPTER 2. THE BCS THEORY
p
-p'
F
p
p'
Figure 2.1: Particle (shaded circle) and hole (white circle) excitations in Landau
Fermi liquid. The particle excitation is obtained by adding a particle. The hole
excitation is obtained by removing a particle (black circle) with an opposite
momentum.
momentum p
and an energy E
from below the Fermi surface, p
 p
F
and E > E
F
, creates an excitation with momentum p and
energy
p
= EE
F
. For an isotropic system, the excitation spectrum will thus
have the form
p
=
_
p
2
2m
E
F
, p > p
F
E
F

p
2
2m
, p < p
F
(2.2)
shown in Fig. 2.2.
The particle and hole excitations live in a system of Fermions where a strong
correlation exists due to the Pauli principle. How well elementary excitations
with a free-particle spectrum Eq. (2.2) are dened here? Consider this con-
dition taking into account the process of quasiparticle-quasiparticle scattering.
Assume that a particle with momentum p
1
above the Fermi surface interacts
with an existing particle p
2
below the Fermi surface after which they both go
to the states p
1
and p
2
above the Fermi surface, see Fig. 2.3. This is the only
possibility according to the Pauli exclusion principle. We thus have
p
1
> p
F
, p
2
 p
F
, p
2
> p
F
(2.3)
This process is shown in Fig. 2.3 together with the momentum conservation
p
1
+p
2
= p
1
+p
2
(2.4)
The probability of such a process is
P = 2m
_
M
_
p
2
1
+p
2
2
p
2
1
p
2
2
_
(n)
(p
1
+p
2
p
1
p
2
) d
n
p
2
d
n
p
1
d
n
p
2
= 2m
_
M
_
p
2
1
+p
2
2
p
2
1
(p
1
+p
2
p
1
)
2
_
d
n
p
2
d
n
p
1
(2.5)
2.1. LANDAU FERMI-LIQUID 25
(p /2m) -E
2
F
p
0
p p
F
Figure 2.2: Single-particle spectrum (p
2
/2m) E
F
(dashed line) is transformed
into the Landau excitation spectrum
p
in a strongly correlated Fermi liquid.
p
p'
p
p'
p + p
1
2
1
2
1 2
Figure 2.3: Scattering of quasiparticles.
The delta functions in the rst line ensure conservations of energy and momenta,
M is the corresponding matrix element, n is the dimension of space.
We shall see that the momenta we are interested in are close to the Fermi
surface. According to the function, this means that p
1
almost coincides with
p
1
and p
2
is close to p
2
. Therefore, Eq. (2.4) gives also
p
1
+p
2
= p
1
+p
2
(2.6)
Indeed, from Eq. (2.4) we nd
p
2
1
+p
2
2
+ 2p
1
p
2
cos( p
1
p
2
) = p
2
1
+p
2
2
+ 2p
1
p
2
cos(

p
1
p
2
)
Since p
1
p
2
p
1
p
2
p
2
F
and p
1
p
2
=

p
1
p
2
we have p
2
1
+p
2
2
= p
2
1
+p
2
2
and also
p
2
1
+p
2
2
+ 2p
1
p
2
= p
2
1
+p
2
2
+ 2p
1
p
2
which leads to Eq. (2.6).
Since p
2
> p
F
, we have
p
1
= p
1
+p
2
p
2
< p
1
+p
2
p
F
Therefore,
0 < p
1
p
F
< (p
1
p
F
) + (p
2
p
F
) (2.7)
On the other hand,
p
F
p
2
< p
1
p
1
< p
1
p
F
thus
0 < p
F
p
2
< p
1
p
F
(2.8)
We can now calculate the integral in Eq. (2.5). Consider rst the 2D and
3D cases, n = 2 or n = 3. In these case, the delta function is removed due to
integration over the angle between p
1
and p
2
. We then have integrals over the
magnitude of momenta. Using Eqs. (2.7) and (2.8) we nd
P
_
dp
2
dp
1
=
_
p
1
p
F
0
d(p
F
p
2
)
_
(p
1
p
F
)+(p
2
p
F
)
0
d(p
1
p
F
)
=
_
p
1
p
F
0
[(p
1
p
F
) (p
F
p
2
)] d(p
F
p
2
) =
1
2
(p
1
p
F
)
2
The coecient of proportionality can be established from the dimensions
P
v
2
F
(p
1
p
F
)
2
hE
F
=

2
hE
F
This means that uncertainty of the quasiparticle energy hP is small com-
pared to the energy if E
F
, i.e., near the Fermi surface. In other words,
quasiparticles are well dened only near the Fermi surface.
For a one dimensional system, however, there is no angular variable, and the
function is removed by integration over dp
1
while the integral over dp
2
gives
of the order of p
F
. Therefore, there is no small parameter in P. Thus, Landau
quasiparticles do not exist. The one-dimensional system of Fermions is known
as the Luttinger liquid, which is beyond the present course.
Let us now dene the Hamiltonian for particles and holes. For particles, we
dene a single-electron Hamiltonian
H
e
=
1
2m
_
ih
e
c
A
_
2
+U
0
(r) (2.9)
where is the chemical potential. In the normal state, = E
F
. Being applied
to a system of totally ^ =
_
ndV particles, it will produce a Hamiltonian in
the form H^ where H is the Hamiltonian of the full system. This is more
appropriate for a system where the chemical potential is xed rather than the
number of particles, as is the case, for example, in superconductors connected
2.2. THE COOPER PROBLEM 27
p'
-p'
p
p
F
-p
p
p
F
(a) (b)
Figure 2.4: (a) A Cooper pair is formed out of a particle excitation with a
momentum p and that with a momentum p above the Fermi surface (shaded
circles), which is equivalent to (b) a particle excitation with a momentum p and
a removed hole excitation (white circle) with nearly the same momentum p
.
to an external circuit. The Hamiltonian Eq. (2.9) corresponds to the canonical
momentum operator

P = ih and is assumed to be spin independent.
The wave function of a particle excitation u
,p
(r) with an energy and
momentum p satises
H
e
u
,p
(r) =
p
u
,p
(r) (2.10)
Let us now turn to hole excitations. The hole wave function v
,p
(r) with an
energy and momentum p satises
H
h
v
,p
(r) =
p
v
,p
(r)
A hole excitation is the absence of a particle with the energy and momentum
p. According to the Landau Fermi-liquid description, the hole Hamiltonian is
thus
H
h
=
e
The Hamiltonian
e
=
1
2m
_
ih
e
c
A
_
2
+U
0
(r) E
F
(2.11)
corresponds to the canonical momentum operator
P = ih. The hole wave

function thus satises
e
v
,p
(r) =
p
v
,p
(r) (2.12)
2.2 The Cooper problem
The original Cooper problem is as follows. Consider an object which is made
out of a pair of electrons with energies
p
having momenta p and p slightly
beyond the Fermi surface, see Fig. 2.4 (a). Their wave functions are u
p
(r) =
e
ipr/ h
U
p
and u
p
(r) = e
ipr/ h
U
p
, respectively. We shall see later that the
pair is actually formed out of an electron in a state u
p
(r) = e
ipr/ h
U
p
above the
Fermi surface and an annihilated hole which was formerly in a state v
p
(r) =
e
ip
r/ h
V
p
with nearly the same momentum p
p below the Fermi surface.

The annihilated hole is in a sense equivalent to an electron with momentum
p
and has a wave function v
p
(r) = u
p
(r) = e
ipr/ h
V
p
, Fig. 2.4 (b).
The pair wave function is
pair
p
(r
1
, r
2
) = u
p
(r
1
)u
p
(r
2
) = e
ip(r
1
r
2
)/ h
U
p
V
p
The linear combination with various p gives the coordinate wave function
pair
(r
1
, r
2
) =
p
e
ip(r
1
r
2
)/ h
a
p
(2.13)
where a
p
= U
p
V
p
. The inverse transformation is
a
p
= V
1
_
(r)e
ipr/ h
d
3
r
where V is the volume of the system.
Assume that the electrons in the pair interact through the potential W(r
1
, r
2
) =
W(r
1
r
2
). Their Hamiltonian is

H
e
(1)+

H
e
(2)+W. The Schr odinger equation
has the form
_

H
e
(r
1
) +

H
e
(r
2
) +W(r
1
, r
2
)
_
pair
(r
1
, r
2
) = E
pair
(r
1
, r
2
)
Multiplying this by e
ip(r
1
r
2
)
and calculating the integral over the volume we
obtain this equation in the momentum representation,
[2
p
E
p
] a
p
=
p
1
W
p,p
1
a
p
1
where
W
p,p
1
= V
1
_
e
i(pp
1
)r/ h
W(r) d
3
r
Assume that
W
p,p
1
=
_
W/V,
p
and
p
1
< E
c
0,
p
or
p
1
> E
c
where E
c
E
F
. The interaction strength W W
0
v
0
where W
0
is the magni-
tude of the interaction potential while v
0
= a
3
0
is the volume where the interac-
tion of a range a
0
is concentrated. We have
a
p
=
W
E 2
p
1
V
p
1
a
p
1
=
W
E 2
p
p
1
a
p
1
(2.14)
Here the sum

p
is taken over p which satisfy
p
< E
c
, while the sum

is
taken over the states in a unit volume. Let us denote
C =
a
p
2.2. THE COOPER PROBLEM 29
Eq. (2.14) yields
a
p
=
WC
E 2
p
whence
C = WC
1
E 2
p
This gives
1
W
=
1
E 2
p
(E) (2.15)
(E)
E
-|W|
-1
2
n
2
0
Figure 2.5: The function (E) for a system with a discrete spectrum
n
.
Equation (2.15) is illustrated in Fig. 2.5. Let us put our system in a large
box. The levels
p
will become a discrete set
n
shown in Fig. 2.5 by vertical
dashed lines. The lowest level
0
is very close to zero and will approach zero
as the size of the box increases. The function (E) varies from to +
as E increases and crosses each
n
> 0. However, for negative E < 0, the
function (E) approaches zero as E , and there is a crossing point with
a negative level 1/[W[ for negative E. This implies that there is a state with
E < 0 satisfying Eq. (2.15) for a negative W < 0.
For an attraction W < 0 we have
1
[W[
=
1
2
p
E
Let n() be the number of states within a unit volume per one spin projection
with energies below . The quantity
N() =
dn()
d
is called the density of states (DOS). In the normal state where
p
= p
2
/2mE
F
,
n() =
(4/3)p
3
(2h)
3
Therefore,
N() =
mp
2
2
h
3
Having this in mind, we substitute the sum with the integral
= 2
_
d
3
p
(2h)
3
= 2
_
mp
2
2
h
3
d
p
the factor 2 accounts for the spin.
Now, for negative energy E = [E
0
[
1
[W[
= 2
_
E
c
0
mp
2
2
h
3
d
p
2
p
E
0
= 2N(0)
_
E
c
0
d
2 +[E
0
[
= N(0) ln
_
[E
0
[ + 2E
c
[E
0
[
_
(2.16)
Here we replace p with a constant p
F
since E
c
E
F
and thus [p p
F
[ p
F
.
We also denote
N(0) =
mp
F
2
2
h
3
(2.17)
the density of states at the Fermi surface. Eq. (2.16) yields
[E
0
[ =
2E
c
e
1/N(0)|W|
1
(2.18)
The dimensionless factor N(0)W N(0)W
0
a
3
0
is called the interaction
constant. For weak coupling, N(0)[W[ 1, we nd
[E
0
[ = 2E
c
e
1/N(0)|W|
For a strong coupling, N(0)[W[ 1,
[E
0
[ = 2N(0)[W[E
c
We see that there exists a state of a particle-hole pair (the Cooper pair) with
an energy [E
0
[ below the Fermi surface. It means that the system of normal-state
particles and holes is unstable towards formation of pairs provided there is an
attraction (however small) between electrons. In conventional superconductors,
the attraction is caused by an exchange of phonons. The attraction between
electrons can also be caused by magnetic interactions which favors triplet pairing
(with a nonzero spin of pair). The Coulomb repulsion is strongly reduced by
screening eects at distances of the order of the size of the pair thus it does
not destroy pairing. Fig. 2.6 illustrates the eect on the excitation spectrum
of coupling between a particle and a hole near the Fermi surface shown in Fig.
2.4.
2.3. THE BCS MODEL 31
E
E
0
p
F
p
Figure 2.6: The coupling between electron and hole modies the energy spec-
trum: The gap equal to [E
0
[ opens near the Fermi surface.
This Cooper pairing eect provides a basis for understanding of supercon-
ductivity. According to this picture, the pairs, being Bose particles, form a Bose
condensate in a single state with a wave function that has a single phase for all
pairs, which is the basic requirement for existence of a spontaneous supercur-
rent.
2.3 The BCS model
2.3.1 Wave functions
According to the above picture, excitations in the normal state are particles and
holes. Using the Landau description of Fermi liquid systems we assume that the
wave functions (annihilation and creation operators) of electrons at a point r
with a spin or in the superconducting state are linear combinations of these
states, i.e., of particles and holes,
(r ) =
n
_
n
u
n
(r)
n
v
n
(r)
_
(2.19)
(r ) =
n
_
n
u
n
(r) +
n
v
n
(r)
_
(2.20)
and
(r ) =
n
_
n
u
n
(r)
n
v
n
(r)
_
(2.21)
(r ) =
n
_
n
u
n
(r) +
n
v
n
(r)
_
(2.22)
Here
n
creates a particle in the state n with the spin up while
n
annihilates
hole with the spin down. The wave functions of excitations u and v are also
modied as compared to the normal state. Our task is to nd the new wave
functions.
The electron operators obey the Fermi commutation rules
(r, )(r
, ) + (r
, )(r, ) = 0
(r, )
(r
, ) +
(r
, )
(r, ) = 0 (2.23)
(r, )(r
, ) + (r
, )
(r, ) =
(r r
)
where is a spin index. The particle and hole operators also obey the Fermi
commutation rules
n,
m,
+
m,
n,
= 0
n,
m,
+
m,
n,
= 0 (2.24)
n,
m,
+
m,
n,
=
mn
To make Eqs. (2.23) and (2.24) compatible the functions u and v should
satisfy the completeness condition
n
[u
n
(r)u
n
(r
) +v
n
(r
)v
n
(r)] = (r r
) (2.25)
and, in addition,
n
[u
n
(r)v
n
(r
) u
n
(r
)v
n
(r)] = 0 (2.26)
To derive Eq. (2.25) we calculate the last anticommutator in Eqs. (2.23).
For example,
(r )(r
) + (r
(r )
=
n,m
_
m
+
m
n
_
u
n
(r)u
m
(r
) +
n,m
_
n
+
n
m
_
v
n
(r)v
m
(r
n,m
[
n
m
+
m
n
] v
n
(r)u
m
(r
n,m
_
m
+
n
_
u
n
(r)v
m
(r
)
=
n
[u
n
(r)u
n
(r
) +v
n
(r)v
n
(r
)] = (r r
)
Here we use Eqs. (2.24). To derive Eq. (2.26), let us calculate the second
anticommutator in Eq. (2.23)
0 =
(r )
(r
) +
(r
(r )
=
m,n
_
m
+
n
_
u
m
(r
)u
n
(r)
m,n
[
n
m
+
m
n
] v
m
(r
)v
n
(r)
+
m,n
_
m
+
m
n
_
u
n
(r)v
m
(r
m,n
_
m
+
n
_
u
m
(r
)v
n
(r)
=
n
[u
n
(r)v
n
(r
) u
n
(r
)v
n
(r)]
If the state is specied by a wave vector q such that u
q
, v
q
e
iqr
, we have
per unit volume
_
d
3
q
(2)
3
and the completeness condition becomes
_
d
3
q
(2)
3
_
u
q
(r)u
q
(r
) +v
q
(r
)v
q
(r)
= (r r
) (2.27)
2.3.2 Hamiltonian
The Hamiltonian operator of the system of electrons has the form
H = H
kin
+H
int
where the kinetic energy operator is
H
kin
=
_
d
3
r
(r, )

H
e
(r, ) (2.28)
and the pairwise point-like interaction is taken in the formW(r
1
, r
2
) = Wr
1
r
2
). Here W W
0
a
3
where W
0
is the magnitude of the interaction while a is
its range. The interaction Hamiltonian is
H
int
=
1
2
,
_
d
3
r
1
d
3
r
2
(r
2
, )
(r
1
, )W(r
1
, r
2
)(r
1
, )(r
2
, )
=
W
2
,
_
d
3
r
(r, )
(r, )(r, )(r, ) (2.29)

It is attractive provided W > 0. The dimension of the pair of operators
is
the dimension of electron density, i.e. L
3
where L is the dimension of length.
Therefore, the dimension of the Hamiltonian is [W]L
6
L
3
= [W
0
]a
3
/L
3
, i.e, it
has the dimension of energy. The total Hamiltonian is of the fourth order in

operators which makes all calculations extremely complicated.

To simplify the theory, we introduce an eective mean-eld Hamiltonian
which is of the second order in

:
H
eff
=
_
d
3
r
(r, )

H
e
(r, ) +U(r)
(r, )(r, )
_
+
_
d
3
r
_
(r)
(r, )
(r, ) +
(r)(r, )(r, ) +H
0
(r)
(2.30)
It contains yet unknown real eective eld U(r) and a complex eective eld
(r). The quantity H
0
(r) does not depend on the elds and
. The idea is
that
(1) This Hamiltonian is diagonal in the new operators
n,
and
n,
,
H
eff
= E
g
+
n,
n,
n,
, (2.31)
where E
g
is the energy of the ground state, while
n
is an energy of the excitation
in a state n.
(2) The eective elds U(r) and (r) are chosen such that the eective
Hamiltonian is as close to the true Hamiltonian as possible. This means that (i)
the statistically averaged H
eff
) has a minimum at the same wave functions as
the average of the true Hamiltonian H). Moreover, (ii) these minimal values
coincide. By denition, the statistical average is
H
eff
) = Z
1
k
_
k
[H
eff
[
k
_
exp(E
k
/T) , (2.32)
where
k
is the set of eigenfunctions of the Hamiltonian
H
eff
k
= E
k
k
The partition function is
Z =
k
exp(E
k
/T) .
Similar equations hold for the true Hamiltonian.
2.3.3 The Bogoliubovde Gennes equations
To full condition (1) let us calculate the commutators
[H
eff
, (r, )]
= H
eff
(r, ) (r, )H
eff
[H
eff
, (r, )]
= H
eff
(r, ) (r, )H
eff
We nd
[H
eff
, (r, )]
=
_

H
e
+U(r)
_
(r, ) (r)
(r, ) (2.33)
[H
eff
, (r, )]
=
_

H
e
+U(r)
_
(r, ) + (r)
(r, ) (2.34)
We now express the operators in both sides of these equations through
the operators using Eqs. (2.21) and (2.22). The commutators [H
eff
,
n,
]
are then calculated using Eq. (2.31):

[H
eff
,
n,
]
=
n
n,
_
H
eff
,
n,
=
n
n,
(2.35)
Comparing the terms with and
we nd that Eqs. (2.35) hold provided

the wave functions u and v satisfy the Bogoliubovde Gennes equations (BdGE)
_

H
e
+U(r)
_
u(r) + (r)v(r) = u(r) (2.36)
_

H
e
+U(r)
_
v(r) +
(r)u(r) = v(r) (2.37)

BdGE can be written in the vector form
_
u
n
v
n
_
=
n
_
u
n
v
n
_
(2.38)
where the matrix operator is
=
_

H
e
+U
[

H
e
+U]
_
The Hermitian conjugated equation is
_
u
n
, v
n
_

+
=
_
u
n
, v
n
_
n
where the transposed operator
+
=
_

H
e
+U
[

H
e
+U]
_
acts on the left. We now multiply Eq. (2.38) by the conjugated vector (u
m
, v
m
)
and integrate over space. Transferring the operators

H
e
and

H
e
to the left we
nd
(
n
m
)
_
[u
m
(r)u
n
(r) +v
m
(r)v
n
(r)] d
3
r = 0
It means that the functions (u, v) are orthogonal
_
[u
m
(r)u
n
(r) +v
m
(r)v
n
(r)] d
3
r =
mn
(2.39)
The factor unity in front of
mn
corresponds to the normalization of the wave
functions Eq. (2.25), (2.27). For the momentum representation we have
_
_
u
q
1
(r)u
q
2
(r) +v
q
1
(r)v
q
2
(r)
d
3
r = (2)
3
(q
1
q
2
) (2.40)
Equations (2.36) and (2.37) posses an important property: If the vector
_
u
n
v
n
_
belongs to an energy
n
(2.41)
then the vector
_
v
n
u
n
_
belongs to the energy
n
. (2.42)
2.3.4 The self-consistency equation
To full condition (2) we write
H) =
_
d
3
r
_
(r, )

H
e
(r, )
_
W
2
,
_
d
3
r
(r, )
(r, )(r, )(r, )

_
We now use the Wicks theorem according to which
(r
1
, )
(r
2
, )(r
3
, )(r
4
, )
_
=

(r
1
, )
(r
2
, )
_
(r
3
, )(r
4
, ))
+
(r
1
, )(r
4
, )
_
(r
2
, )(r
3
, )
_
(r
1
, )(r
3
, )
_
(r
2
, )(r
4
, )
_
Therefore
(r, )
(r, )(r, )(r, )

_
=

(r, )
(r, )
_
(r, )(r, ))
+
(r, )
(r, )
_
(r, )(r, ))
+
(r, )(r, )
_
(r, )(r, )
_
(r, )(r, )
_
(r, )(r, )
_
Here we assume that
(r, )
(r, )
_
,= 0 and (r, )(r, )) ,= 0
due to the Cooper pairing, while
(r, )
(r, )
_
= 0 and (r, )(r, )) = 0
because pairing occurs for particles with opposite spins.
Assume also that
(r, )(r, )
_
= 0 and

(r, )(r, )
_
= 0
due to the absence of magnetic interaction. Using the anticommutation relations
we nally obtain
(r, )
(r, )(r, )(r, )

_
= 2
(r, )
(r, )
_
(r, )(r, ))
+
(r, )(r, )
_
(r, )(r, )
_
(r, )(r, )
_
(r, )(r, )
_
= 2
(r, )
(r, )
_
(r, )(r, ))
+2
(r, )(r, )
_
(r, )(r, )
_
The variation of the true energy becomes
H) =
_
d
3
r
_
(r, )

H
e
(r, )
_
W
_
d
3
r
__
(r, )
(r, )
__
(r, )(r, ))
+
(r, )
(r, )
_
( (r, )(r, )))
W
_
d
3
r
__
(r, )(r, )
__
(r, )(r, )
_
+

(r, )(r, )
_ _
(r, )(r, )
__
(2.43)
Let us now calculate the variation of the eective energy with respect to
variations of :
H
eff
) =
_
d
3
r
_
(r, )
_

H
e
+U(r)
_
(r, )
_
+
_
d
3
r
_
(r)
(r, )
(r, )
_
+
(r) (r, )(r, ))
(2.44)
Comparing the corresponding terms in Eqs. (2.43) and (2.44) we nd that
there should be
U(r) = W
(r, )(r, )
_
= W
(r, )(r, )
_
(2.45)
and
(r) = W (r, )(r, )) = W (r, )(r, )) (2.46)
(r) = W
(r, )
(r, )
_
= W
(r, )
(r, )
_
(2.47)
The pairing eld is proportional to a two-particle wave function.
To calculate the averages in Eqs. (2.45) (2.47) we use
n,
m,
_
=
nm
f
n
(2.48)
n,
m,
) =
_
n,
m,
_
= 0 (2.49)
where f
n
is the distribution function. In equilibrium, it is the Fermi function
f
n
=
1
e
n
/T
+ 1
Eqs. (2.48) and (2.49) reect the fact that are the single-particle operators in
the new basis such that there is no further pairing among them.
Inserting Eqs. (2.19)(2.22) into Eqs. (2.46), (2.47) we nd
(r) = W
n
(1 f
n
f
n
) u
n
(r)v
n
(r) = W
n
(1 2f
n
) u
n
(r)v
n
(r) (2.50)
(r) = W
n
(1 f
n
f
n
) u
n
(r)v
n
(r) = W
n
(1 2f
n
) u
n
(r)v
n
(r) (2.51)
The last equality in each line holds because the distribution is spin independent.
We see that the pairing eld is a linear combination of pair states made out
of particle-like and annihilated hole-like excitations.
The eective potential is
U(r) = W
n
_
[u
n
(r)[
2
f
n
+[v
n
(r)[
2
(1 f
n
)
(2.52)
Note that the density of particles in the system is
n = 2
(r, )(r, )
_
= 2
n
_
[u
n
(r)[
2
f
n
+[v
n
(r)[
2
(1 f
n
)
(2.53)
One can say that it is a combination of particle contribution
2
n
[u
n
(r)[
2
f
n
and a hole contribution
2
n
[v
n
(r)[
2
(1 f
n
)
where the distribution of holes is 1 f
n
.
The eective eld U enters the BdGE together with the chemical potential
. In a sense, it can be included into to make a new chemical potential. In
what follows we assume that it is included into and omit it from the BdGE.
We can now write the average energy in the state satisfying Eqs. (2.46) ,
(2.47) with the eective eld as
H)
=
_
d
3
r
_
(r, )

H
e
(r, )
_
[(r)[
2
W
_
(2.54)
The average eective Hamiltonian Eq. (2.30) in the state determined by Eqs.
(2.46), (2.47) coincides with the true average energy in this state if H
0
(r) =
[[
2
/V . Therefore, the full expression is
H
eff
=
_
d
3
r
_
(r, )

H
e
(r, ) +
[(r)[
2
W
_
+
_
d
3
r
_
(r)
(r, )
(r, ) +
(r)(r, )(r, )
(2.55)
With the potential U included into the chemical potential, the BdG equa-
tions take the form
h
2
2m
_
ie
hc
A
_
2
u E
F
u + v = u (2.56)
h
2
2m
_
+
ie
hc
A
_
2
v +E
F
v +
u = v (2.57)
2.4. OBSERVABLES 39
2.4 Observables
2.4.1 Energy spectrum and coherence factors
Consider the case where = [[e
i
is constant in space, and the magnetic eld
is absent. The BdGE have the form
_
h
2
2m
_
u(r) + v(r) = u(r) (2.58)
h
2
2m
_
v(r) +
u(r) = v(r) (2.59)

where = h
2
k
2
F
/2m. We look for a solution in the form
u = e
i
2
U
q
e
iqr
, v = e
i
2
V
q
e
iqr
(2.60)
where q is a constant vector. We have
q
U
q
+[[V
q
=
q
U
q
(2.61)
q
V
q
+[[U
q
=
q
V
q
(2.62)
where
q
=
h
2
2m
_
q
2
k
2
F
The condition of solvability of Eqs. (2.61) and (2.62) gives
q
=
_
2
q
+[[
2
(2.63)
According to the Landau picture of Fermi liquid, we consider only energies > 0.
The spectrum is shown in Fig. 2.7.
The wave functions u and v for a given momentum q are found from Eqs.
(2.61), (2.63). We have
U
q
=
1
2
_
1 +

q
q
_
1/2
, V
q
=
1
2
_
1

q
q
_
1/2
(2.64)
Normalization is chosen to satisfy Eq. (2.27).
The energy [[ is the lowest single-particle excitation energy in the super-
conducting state. 2[[ corresponds to an energy which is needed to destroy
the Cooper pair. Therefore, one can identify 2[[ as the pairing energy as
determined by the Cooper problem in the previous Section, [E[ = 2[[. Us-
ing the results of Problem 2.1, one can dene the radius of the Cooper pair
_
R
2
) hv
F
/[[. This characteristic scale sets up a very important length
scale called the coherence length
hv
F
/[[ .
E
p
p
F
F
particles holes
Figure 2.7: The BCS spectrum of excitations in a superconductor. The solid

line shows the spectrum of quasiparticles near the Fermi surface where the
Landau quasiparticles are well dened. At higher energies closer to E
F
the
Landau quasiparticles are not well-dened (dotted line). The dashed line at
lower energies shows the behavior of the spectrum in the normal state = [
p
[.
For a given energy, there are two possible values of
q
=
_
2
[[
2
(2.65)
that correspond respectively to particles or holes (see Fig. 2.7).
The quantity
d
d( hq)
= v
g
is the group velocity of excitations. One has q
2
= k
2
F
(2m/h
2
)
_
2
[[
2
.
Therefore,
v
g
=
hq
m
q
_
2
q
+[[
2
= v
F
q
_
2
q
+[[
2
= v
F
_
2
[[
2
(2.66)
where v
F
= hq/m is the velocity at the Fermi surface. We see that the group
velocity is positive, i.e., its direction coincides with the direction of q for ex-
citations outside the Fermi surface
q
> 0. As we know, these excitations are
particle-like. On the other hand, v
g
< 0 for excitations inside the Fermi surface
q
< 0, which are known as hole-like excitations.
Another important quantity is the density of states (DOS) N() dened as
follows. Let us suppose that there are n
(q) states per spin and per unit volume

for particles with momenta up to hq. The density of states is the number of
states within an energy interval from to +d, i.e.
N() =
dn
(q)
d
2.4. OBSERVABLES 41
As we can see, in a superconductor, there are no excitations with energies <
[[. The DOS per one spin projection is zero for < [[. For > [[ we have
N() =
1
2
d
d
_
q
3
3
2
_
=
q
2
2
2
d
dq
1
=
mq
2
2
h
2
2
[[
2
N(0)

_
2
[[
2
(2.67)
Here
N(0) =
mp
F
2
2
h
3
=
mk
F
2
2
h
2
is the DOS per one spin projection in the normal state for zero energy excita-
tions, i.e., at the Fermi surface. We have replaced here q with k
F
. Indeed, since
E
F
, the magnitude of q is very close to k
F
for . This fact is of a
crucial importance for practical applications of the BCS theory, as we shall see
in what follows.
One nds from Eq. (2.65) that for a given energy there are two possible
values of q
2
:
q
2
= k
2
F
+
2m
h
2

q
= k
2
F

2m
h
2
_
2
[[
2
(2.68)
Using this we can write Eq. (2.68) in the form
q = q
where
q
= k
F

1
hv
F
_
2
[[
2
(2.69)
Therefore, the solution Eq. (2.60) for a given energy consists of four terms
_
u
v
_
= A
_
e
i
2
U
0
e
i
2
V
0
_
e
iq
+
x
+B
_
e
i
2
V
0
e
i
2
U
0
_
e
iq
x
+C
_
e
i
2
U
0
e
i
2
V
0
_
e
iq
+
x
+D
_
e
i
2
V
0
e
i
2
U
0
_
e
iq
x
(2.70)
Here we have chosen the x axis parallel to q. The amplitudes A and C describe
excitations that have the same directions of the momenta q
+
and of the group
velocity v
g
, the amplitudes B and D describe excitations that have opposite
directions of the momenta q
and of the group velocity v

g
.
The relations between U
0
and V
0
can be found if we insert each term in Eq.
(2.70) into Eq. (2.61) or Eq. (2.62). We nd
U
0
=
1
2
_
1 +
_
2
[[
2
_
1/2
, V
0
=
1
2
_
1
_
2
[[
2
_
1/2
(2.71)
The normalization is chosen such that each set of U and V satises Eq. (2.27)
separately. The amplitudes U and V are sometimes called the coherence factors.
In the normal state = 0 the wave function for a given energy becomes
_
u
v
_
= A
_
1
0
_
e
iq
+
x
+B
_
0
1
_
e
iq
x
+C
_
1
0
_
e
iq
+
x
+D
_
0
1
_
e
iq
x
(2.72)
We thus have particle waves with the group velocity parallel to the momentum
u = Ae
iq
+
x
+Ce
iq
+
x
(2.73)
and, independently, the holes waves with the group velocity antiparallel to the
momentum
v = Be
iq
x
+De
iq
x
(2.74)
2.4.2 The energy gap
For a state Eq. (2.60) specied by a wave vector q with U
q
and V
q
from Eq.
(2.64) the product uv
becomes
uv
= e
i
U
q
V
q
=

2
q
The self-consistency equation (2.50) yields
= W
q
(1 2f
q
)

2
q
(2.75)
We replace the sum with the integral
q
=
_
q=
q=0
d
3
q
(2)
3
=
_
=+
=E
F
mq
2
2
h
2
d
q
N(0)
_
+
d
q
and notice that, in equilibrium,
1 2f
q
= tanh
_

q
2T
_
Moreover,
q
d
q
=
q
d
q
When varies from to +, the energy varies from to + taking each
value twice. Therefore, the self-consistency equation takes the form
= W
q
(1 2f
q
)

2
q
= N(0)W
_

||
2
[[
2
tanh
_

2T
_
d (2.76)
The integral diverges logarithmically at large energies. In fact, the potential
of interaction does also depend on energy W = W
in such a way that it vanishes

for high energies exceeding some limiting value E
c
E
F
. We assume that
W
=
_
W, < E
c
0, > E
c
2.4. OBSERVABLES 43
From Eq. (2.76) we obtain the gap equation
1 =
_
E
c
||
1
_
2
[[
2
tanh
_

2T
_
d (2.77)
where the dimensionless parameter = N(0)W N(0)W
0
a
3
is called the inter-
action constant. For phonon mediated electron coupling, the eective attraction
works for energies below the Debye energy
D
. Therefore, E
c
=
D
in Eq.
(2.77). This equation determines the dependence of the gap on temperature.
This equation can be used to determine the critical temperature T
c
, at which
the gap vanishes. We have
1 =
_
E
c
0
tanh
_

2T
c
_
d
(2.78)
This reduces to
1
=
_
E
c
/2T
c
0
tanh x
x
dx (2.79)
The integral
_
a
0
tanh x
x
dx = ln(aB)
Here B = 4/ 2.26 where = e
C
1.78 and C = 0.577 . . . is the Euler
constant. Therefore,
T
c
= (2/)E
c
e
1/
1.13E
c
e
1/
(2.80)
The interaction constant is usually small, being of the order of 0.1 0.3 in
practical superconductors. Therefore, usually T
c
E
c
.
For zero temperature we obtain from Eq. (2.77)
1
=
_
E
c
||
d
_
2
[[
2
= Arcosh
_
E
c
[[
_
ln(2E
c
/[[) (2.81)
Therefore, at T = 0
[[ (0) = (/)T
c
1.76T
c
2.4.3 Condensation energy
Let us calculate the average energy per unit volume in the superconducting
state, Eq. (2.54). Using Eqs. (2.19 - 2.22) and (2.48), (2.49) we have
_
d
3
r
(r, )

H
e
(r, )
_
=
n,m,
_
n
_
__
u
m

H
e
u
n
d
3
r
_
+
n
_
__
v
m

H
e
v
n
d
3
r
__
= 2
n
_
f
n
__
u
n

H
e
u
n
d
3
r
_
+ (1 f
n
)
__
v
n

H
e
v
n
d
3
r
__
Using the BdGE and the normalization of u and v we obtain
_
d
3
r
(r, )

H
e
(r, )
_
= 2
n
_
n
_
f
n
[u
n
[
2
(1 f
n
)[v
n
[
2
_
+ v
n
u
n
(1 2f
n
)
= 2
n
_
n
f
n
n
[v
n
[
2
+
2[[
2
W
Therefore,
H)
= 2
n
f
n
2
n
[v
n
[
2
+
[[
2
W
(2.82)
Let us consider zero temperature. The rst term gives the energy of excita-
tions. This energy is zero for T = 0, since f
n
= 0 for any
n
> 0. The rest two
terms determine the ground state energy measured from the energy in normal
state
c
= 2
n
[v
n
[
2
+
[[
2
W
=
n
(
n
n
) +
[[
2
W
= 2N(0)
_
E
c
0
_
_
2
+[[
2
_
d +
[[
2
W
We have
_
E
c
0
_
_
2
+[[
2
_
d =
1
2
_
2
+[[
2
2
+[[
2
ln( +
_
2
+[[
2
)
_
E
c
0
=
[[
2
4
+
[[
2
2
ln
_
2E
c
[[
_
=
[[
2
4
+
[[
2
2N(0)W
Here we use that [ E
c
and insert the logarithm from Eq. (2.81). We nd
c
=
N(0)[(0)[
2
2
(2.83)
The ground state energy of the superconductor is by an amount c
lower
than the normal state energy. The energy c
is called the condensation energy.

2.4.4 Current
The quantum mechanical expression for the current density is
j =
e
2m
(r, )
__
ih
e
c
A
_
(r, )
_
+
__
ih
e
c
A
_
(r, )
_
(r, )
_
(2.84)
2.4. OBSERVABLES 45
Using Eqs. (2.19 - 2.22) and (2.48), (2.49) we obtain
j =
e
m
n
_
f
n
u
n
(r)
_
ih
e
c
A
_
u
n
(r)
+ (1 f
n
) v
n
(r)
_
ih
e
c
A
_
v
n
(r) +c.c.
_
(2.85)
Identities
One can write Eq. (2.85) in the form
j =
ihe
m
n
__
1
2
+f
n,
_
u
n
_
ie
hc
A
_
u
n
+
_
1
2
f
n,
_
v
n
_
ie
hc
A
_
c.c
_
ihe
2m
n
_
u
n
_
ie
hc
A
_
u
n
+v
n
_
ie
hc
A
_
v
n
c.c
_
=
ihe
2m
n
_
(1 2f
n,
) u
n
_
ie
hc
A
_
u
n
(1 2f
n,
) v
n
_
ie
hc
A
_
v
n
c.c
_
ihe
2m
n
_
u
n
_
ie
hc
A
_
u
n
+v
n
_
ie
hc
A
_
v
n
c.c
_
(2.86)
Therefore,
j =
n
(1 2f
n
) J
n
+
n
P
n
(2.87)
where
J
n
=
ihe
2m
_
u
n
_
ie
hc
A
_
u
n
v
n
_
ie
hc
A
_
v
n
c.c
_
(2.88)
and
P
n
=
ihe
2m
_
u
n
_
ie
hc
A
_
u
n
+v
n
_
ie
hc
A
_
v
n
c.c
_
. (2.89)
For a given state n, the current J
n
is not conserved, see below. On the other
hand, P
n
is a conserved quantity (see Problem 2.2.), the quasiparticle ow,
divP
n
= 0 (2.90)
Now, we multiply Eq. (2.56) by u
and subtract the complex conjugated

equation. Next, we multiply Eq. (2.57) by v
and subtract the complex conju-

gated equation. Finally, subtracting the two resulting equations we nd
div J
n
=
2ie
h
(u
n
v
n
u
n
v
n
) . (2.91)
Applying Eqs. (2.87) and (2.90) we can calculate the divergence of the total
current
div j =
2ie
h
_
n
u
n
v
n
(1 2f
n
)
n
u
n
v
n
(1 2f
n
)
_
= 0 (2.92)
due to the self-consistency equation. The sum here is taken over all states. The
total current is conserved as it should be.
Currents in clean superconductors
If the vector potential varies slowly, one can use a gauge where it is included
into k. Assume that the order parameter has a superconducting momentum
= [[e
ikr
The wave functions have the form
u(r) = e
i(q+k/2)r
U
q
, v(r) = e
i(qk/2)r
V
q
(2.93)
The energy spectrum is (see Problem 2.2)
q
= hq v
s
+
_
2
q
+[[
2
(2.94)
or
q
= hq v
s
+
(0)
q
(2.95)
where
(0)
q
=
_
2
q
+[[
2
and
v
s
=
hk
2m
is the superconducting velocity. The factors U
q
and V
q
are given by
U
q
=
1
2
_
1 +

q
(0)
q
_
1/2
, V
q
=
1
2
_
1

q
(0)
q
_
1/2
(2.96)
similar to Eq. (2.64).
Note that the gap vanishes for excitations with q antiparallel to v
s
if v
s
v
c
where
v
c
= [[/p
F
is called the critical velocity.
Using Eqs. (2.64) and (2.93) we obtain from Eq. (2.85)
j =
2he
m
q
__
q +
k
2
_
f
q
U
2
q
_
q
k
2
_
(1 f
q
)V
2
q
_
=
he
m
k
q
_
f
q
U
2
q
+ (1 f
q
)V
2
q
+
2he
m
q
q
_
f
q
U
2
q
(1 f
q
)V
2
q
The distribution function here is

f
q
f(
q
) =
1
e
q
/T
+ 1
2.4. OBSERVABLES 47
where
q
is determined by Eq. (2.94). We can write
j =
he
m
k
q
_
f(
(0)
q
)U
2
q
+ (1 f(
(0)
q
))V
2
q
_
+
2he
m
q
q
_
f(
(0)
q
)U
2
q
(1 f(
(0)
q
))V
2
q
_
+
he
m
k
q
_
f(
q
) f(
(0)
q
)
_
_
U
2
q
V
2
q
+
2he
m
q
q
_
f(
q
) f(
(0)
q
)
_
_
U
2
q
+V
2
q
(2.97)
The rst term in Eq. (2.97) gives
j
0
=
hen
2m
k = nev
s
where particle density n is given by Eq. (2.53). The current j
0
corresponds to
the ow of all particles with the velocity v
s
. The second term in Eq. (2.97)
vanishes after summation over directions of q. The third term also vanishes
since U
2
q
V
2
q
is an odd function of
q
. The current becomes j = j
0
j
norm
where
j
norm
=
2he
m
q
q
_
f(
(0)
q
) f(
q
)
_
is determined by the fourth term. This is the current produced by excitations
which form the normal component of the superconductor. We have
f(
(0)
q
) f(
q
) =
1
e
2
q
+||
2
/T
+ 1
1
e
(
2
q
+||
2
+ hqv
s
)/T
+ 1
The normal current is thus directed along v
s
. We will obtain
j
norm
= en
norm
v
s
where n
norm
is the density of the normal component, which generally depends
on v
s
. The total current is
j = e(n n
norm
)v
s
= en
s
v
s
where n
s
= n n
norm
is the density of superconducting electrons. It also
generally depends on v
s
.
For small v
s
we have a linear dependence
j
norm
=
2h
2
e
m
q
q(q v
s
)
df(
(0)
q
)
d
(0)
q
=
2h
2
e
3m
v
s
q
q
2
df(
(0)
q
)
d
(0)
q
Therefore,
n
norm
=
2p
2
F
3m
q
df(
(0)
q
)
d
(0)
q
=
2p
2
F
3m
N(0)
_

df(
(0)
q
)
d
(0)
q
d
q
= n
_

df(
(0)
q
)
d
(0)
q
d
q
= 2n
_

||
2
[[
2
df()
d
d (2.98)
For = 0 we have n
norm
= n. The normal density ^
norm
is exponentially
small for T [[.
2.4.5 Negative energies
Owing to the property of the BdG equations expressed by Eqs. (2.41) and (2.42),
one can introduce into consideration negative energies of excitations. This also
complies with the negative sign of Eq. (2.63). With the negative energies, the
self-consistency equation (2.50) takes the form
= V
n
(1 2f
n
) u
n
(
n
)v
n
(
n
)
=
V
2
n,>0
[[1 2f(
n
)] u
n
(
n
)v
n
(
n
) [1 2f(
n
)] v
n
(
n
)u
n
(
n
)]
=
V
2
n, >0
[1 2f(
n
)] u
n
(
n
)v
n
(
n
)
V
2
n, <0
[1 2f(
n
)] u
n
(
n
)v
n
(
n
)
In the second term we made a change . In equilibrium,
1 2f() = tanh

2T
is an odd function of . We have
=
V
2
n,all
[1 2f()]

2
n
or
= V
n,>0
2
tanh

2T
(2.99)
which coincides with Eq. (2.76).
2.5 Impurities. Anderson theorem
So far we considered a pure material that has no impurities. Generally, impuri-
ties aect all the properties of superconductors. However, for a superconductor
with nonmagnetic impurities in a zero magnetic eld that does not carry cur-
rent, the critical temperature and the energy gap are the same as in the clean
host material. This is the contents of the Anderson theorem [9].
2.5. IMPURITIES. ANDERSON THEOREM 49
To see this we consider the BdG equations
_
h
2
2m
2
+U(r)
_
u(r) + v(r) = u(r)
h
2
2m
2
+U(r)
_
v(r) +
u(r) = v(r)
Here U includes the impurity potential. The gap does not depend on coordi-
nates since there is no magnetic eld and current. Consider the single-electron
functions of the normal metal w
n
(r) that satisfy
_
h
2
2m
2
+U(r)
_
w
n
(r) =
n
w
n
(r)
Since = const we have obviously
u
n
(r) = w
n
(r)U
n
v
n
(r) = w
n
(r)V
n
where U
n
and V
n
satisfy
(
n
n
) U
n
+ V
n
= 0
(
n
+
n
) V
n
+
U
n
= 0
This gives the usual energy spectrum
n
=
_
2
n
+[[
2
and the coherence factors
[U
n
[
2
=
1
2
_
1 +

n
n
_
, [V
n
[
2
=
1
2
_
1

n
n
_
The self-consistency equation becomes
= W
n
[w
n
(r)[
2
(1 2f
n
)

2
n
We also use the denition of the density of states
N
(r) =
n
[w
n
(r)[
2
(
n
)
The self-consistency equation takes the form
(r) = W
_
N
(r)d (1 2f[()])

2()
The wave functions of the normal metal w
n
(r) oscillate as functions of r
over atomic distances. The squared [w
n
(r)[
2
averages into a spatially uniform
function if one looks at it at distances of the order of the coherence length
essential for superconductivity. Moreover, E
F
. Therefore, after averaging
we nally nd
= WN
0
_
d (1 2f())

2
where N
0
is the average density of states in the normal metal at the Fermi
energy which coincides with that in the clean host material. It is clear that the
gap and thus the critical temperature are the same as in the clean material.
The superconducting and normal density, however, are strongly modied by
the impurities because they are dened for a current-carrying state.
2.5. IMPURITIES. ANDERSON THEOREM 51
Problems
Problem 2.1.
Calculate the mean square radius of the Cooper pair.
Problem 2.2.
Derive equations (2.90) where P is dened by Eq. (2.89).
Problem 2.3.
Derive equations (2.91) where J is dened by Eq. (2.88).
Problem 2.4.
Find the energy spectrum and the coherence factors for the order parameter
= [[e
ikr
.
Problem 2.5.
Derive the gap equation that determines the dependence of [[ on v
s
for the
order parameter in the form = [[e
ikr
.
Problem 2.6
Find the temperature dependence of the gap for T T
c
.
Problem 2.7
Find the temperature dependence of the normal density for T 0.
Problem 2.8
Find the temperature dependence of the superconducting density for T
T
c
.
Chapter 3
Andreev reection
3.1 Semiclassical approximation
3.1.1 Andreev equations
We assume that there are no barriers or other potentials that vary over distances
of the order of the electronic wave length k
1
F
. The energy gap should also be
smooth over the distances of the order of the electronic wave length. It may,
however, vary at distances shorter than the coherence length . In this case the
quasiparticle momentum of the order of hk
F
is a good quantum number such
that we can look for a solution in the form
_
u
v
_
= e
ikr
_
U(x)
V (x)
_
(3.1)
where [k[ = k
F
while U(x) and V (x) vary slowly over distances of the order of
k
1
F
. Inserting this into the BdG equations Eqs. (2.56) and (2.57)
h
2
2m
_
ie
hc
A
_
2
u
h
2
k
2
F
2m
u + v = u (3.2)
h
2
2m
_
+
ie
hc
A
_
2
v +
h
2
k
2
F
2m
v +
u = v (3.3)
and neglecting the second derivatives of U and V , we nd
ihv
F

_
ie
hc
A
_
U + V = U (3.4)
ihv
F

_
+
ie
hc
A
_
V +
U = V (3.5)
The transformation we use is known as the WKB or semiclassical approx-
imation. It is essentially based on the fact that the electronic wave length is
the shortest length scale in superconductors. The new equations are called An-
dreev equations. Each of them is a rst-order equation, which may considerably
reduce the complexity of the problem.
53
54 CHAPTER 3. ANDREEV REFLECTION
3.1.2 Andreev reection
Consider a particle incident from the normal region on the superconducting
half-space x > 0 (see Fig. 3.1) with a momentum parallel to the x axis. Assume
that the gap varies over distances longer than the electron wave length k
1
F
and that both the normal metal and the superconductor have the same Fermi
velocity, and there are no insulating barriers between them. We assume that
the magnetic eld is absent. In this case and all other potentials depend only
on one coordinate x and the Andreev equations take the form
ihv
x
dU
dx
+ V = U (3.6)
ihv
x
dV
dx
+
U = V (3.7)
e
h
e
x 0
Figure 3.1: The NS structure.

Consider rst the case of high energies > [[. The particle will penetrate
into the superconductor and partially will be reected back into the normal
metal. However, since the gap varies slowly, the reection process where the
momentum is changed such that k k is prohibited. We thus can use the
semiclassical approximation.
p
F
p
F
-p
F
-p
F

N
0
S
0
i a c
Figure 3.2: The Andreev reection.
3.1. SEMICLASSICAL APPROXIMATION 55
In the normal region (on the left) the order parameter decreases to zero at
distances shorter than , so that one can consider = 0 for x < 0. Therefore
the functions U(x) and V (x) are independent for x < 0. We look for the wave
function in the form
_
U(x)
V (x)
_
L
= e
i
N
x
_
1
0
_
+ae
i
N
x
_
0
1
_
(3.8)
We nd from Eqs. (3.6), (3.7)
N
=

hv
x
(3.9)
The wave function thus contains an incident particle [state (i) in Fig. 3.2] and a
reected hole [state (a)]. We choose the coecient unity in front of the incident
wave to ensure the unit density of particles in the incident wave.
The wave function on the right (in the S region) has the form
_
U(x)
V (x)
_
R
= ce
i
S
x
_
U
0
V
0
_
(3.10)
Eqs. (3.4), (3.5) give
S
=
_
2
[[
2
hv
x
(3.11)
It describes a transmitted particle. The coherence factors U
0
and V
0
are deter-
mined by Eq. (2.71). They satisfy
U
2
0
V
2
0
=
_
2
[[
2
, U
0
V
0
=
[[
2
The boundary conditions for rst-order dierential equations require conti-
nuity of the functions at the interface whence
a = V
0
/U
0
, c = 1/U
0
(3.12)
The coecient a describes a process when a particle is reected as a hole from
a spatially non-uniform gap; this process is called the Andreev reection [10].
Using Eq. (3.12), the balance of the probabilities in the semi-classical ap-
proximation gives
[a[
2
+ (U
2
0
V
2
0
)[c[
2
= 1
or
v
g
(N)[a[
2
+v
g
(S)[c[
2
= v
g
(N) (3.13)
where, as in Eq. (2.66),
v
g
(S) = v
F
(U
2
0
V
2
0
) = v
F
_
2
[[
2

is the group velocity in the superconducting state, and v
g
(N) = v
F
is the group
velocity in the normal state. Eq. (3.13) complies with the conservation of the
quasiparticle ow P determined by Eq. (2.89).
For the sub-gap energy < [[, there are no states below the gap in the S
region, thus the wave should decay for x > 0. The wave function on the right is
_
U(x)
V (x)
_
R
= ce
i
S
x
_

U
0
V
0
_
(3.14)
where
S
= i
_
[[
2
2
hv
x
(3.15)
and
U
0
=
1
2
_
1 +i
_
[[
2
_
,

V
0
=
1
2
_
1 i
_
[[
2
_
The coecients are
a =

V
0
/
U
0
, c = 1/
U
0
However, now
[a[
2
= 1 (3.16)
The Andreev reection is complete since there are no transmitted particles.
annihilated hole
penetrating particle incident particle
reflected hole
p
p p
p
v
g
Cooper
pair
N S
-|e|
+|e|
-2|e|
v
g
-2|e|
Figure 3.3: Illustration of the nature of Andreev reection for < [ at an ideal
SN interface: An incident electron forms a Cooper pair in the superconductor
together with an annihilated hole. This hole is expelled into the normal metal
and moves back as a reected object.
The Andreev reection has an interesting and surprising property. Consider
a three-dimensional motion. Eq. (3.8) for the normal region gives
_
u
v
_
= e
i(k
x
+/ hv
x
)x+ik
y
y+ik
z
z
_
1
0
_
+ae
i(k
x
/ hv
x
)x+ik
y
y+ik
z
z
_
0
1
_
3.1. SEMICLASSICAL APPROXIMATION 57
q
+
q
h
g
v
g
v
e
x
y
S N
Figure 3.4: The trajectories of an incident particle and the Andreev reected
hole.
The magnitude squared of the particle momentum is
p
2
= p
2
x
+p
2
y
+p
2
z
= h
2
[(k
x
/hv
x
)
2
+k
2
y
+k
2
z
] = h
2
_
k
2
F

2k
x
hv
x
_
= h
2
k
2
F
2m
The total momentum of the incident particle is thus p = hk
F
+/v
F
such that
p > p
F
. Its energy corresponds to the rising (right) part of the spectrum branch
[point (i)] in Fig. 3.2,
(p) = v
F
(p p
F
)
The reected hole has the momentum p = p
F
/v
F
such that p < p
F
. Its
energy thus corresponds to the decreasing (left) part of the spectrum branch
[point (a)] in Fig. 3.2,
(p) = v
F
(p
F
p)
One can calculate the components of the group velocity
v
g x
=

p
x
=
p
x
m
v
x
v
g y
=

p
y
=
p
y
m
v
y
We see that particle and hole have opposite signs of the group velocity but with
almost the same magnitude. Therefore, Andreev reected hole moves along the
same trajectory as the incident particle but in the opposite direction!
In fact, directions of the incident and reected trajectories are slightly dier-
ent (see Problem 3.1). Indeed, the change in the momentum during the Andreev
reection is
p
x
= ( hk
x
/v
x
) ( hk
x
+/v
x
) = 2/v
x
This change is much smaller than the momentum itself. This is because the
energy of interaction is much smaller than E
F
. At the same time, the
momentum projections p
y
and p
z
are conserved. As a result, the trajectory of
the reected hole deviates, but only slightly, from the trajectory of the incident
particle (see Fig. 3.4).
3.2 Andreev states
3.2.1 SNS structures
Consider an SNS structure consisting of a normal slab in the plane (y, z) having
a thickness d in between two superconducting half-spaces x < d/2 and x > d/2
(see Fig. 3.5). The electrons in both superconductors and in the normal metal
have the same Fermi velocity, and there are no insulating barriers between
them. As before, we assume that electrons do not scatter on impurities or on
other objects like phonons. We also assume that there is no electronelectron
scattering. This means that the size d is shorter than the electron mean free
path. For short d the mean free path should be larger than the superconducting
coherence length or
N
(see below).
We use the semi-classical approximation and look for the wave function in
the from
_
u
v
_
= e
ikr
_
U(x)
V (x)
_
e
h
x
0
-d/2 d/2
Figure 3.5: The Andreev bound states in the SNS structure.
Consider energy < [[. A particle that moves in the normal region to the
right will be Andreev reected from the NS interface into a hole. The hole will
then move to the left and is Andreev reected into the particle, and so on. It is
thus localized in the N region having a discrete energy spectrum.
Assume that the gap in the right superconductor has the phase /2 while
the phase in the left superconductor is /2 so that the phase dierence is .
The wave function in the N region is
_
U(x)
V (x)
_
N
= A
_
e
i
N
x
_
1
0
_
+ae
i
N
x
_
0
1
__
(3.17)
3.2. ANDREEV STATES 59
where
N
=

hv
x
If the momentum projection on the x axis is positive, k
x
> 0, the wave
function in the right superconductor x > d/2 is
_
U(x)
V (x)
_
R
= d
1
e
S
x
_

Ue
i/4
V e
i/4
_
(3.18)
where
S
=
_
[[
2
2
h[v
x
[
The wave function in the left superconductor x < d/2 is
_
U(x)
V (x)
_
L
= d
1
e
S
x
_

V e
i/4
Ue
i/4
_
(3.19)
Applying continuity at the right interface x = d/2 we nd
Ae
i
N
d/2
= d
1
e
S
d/2

Ue
i/4
Ae
i
N
d/2
= d
1
e
S
d/2
V e
i/4
whence
ae
i
N
d
=

V
U
e
i/2
Continuity at the left interface gives
ae
i
N
d
=

U
V
e
i/2
Excluding a we nd
e
2i(
N
d/2)
=
+i
_
[[
2
2
i
_
[[
2
2
We denote
sin =

[[
The range of is determined in such a way that
cos = [[
1
_
[[
2
2
is positive to ensure the decay of wave functions in the N regions, i.e.,
/2 < < /2
We obtain
e
2i(
N
d/2)
= e
2i+i
or
= h
x
_
2
arcsin

[[
+
_
l +
1
2
__
(3.20)
where l is an integer and
x
=
v
x
d
= t
1
x
is the inverse time needed for a particle to y from one end of the normal region
to the other.
Consider now particles that have a negative projection k
x
< 0 on the x axis.
The wave function in the right superconductor x > d/2 becomes
_
U(x)
V (x)
_
R
= d
2
e
S
x
_

V e
i/4
Ue
i/4
_
(3.21)
The wave function in the left superconductor x < d/2 is
_
U(x)
V (x)
_
L
= d
2
e
S
x
_

Ue
i/4
V e
i/4
_
(3.22)
Applying continuity at the right interface x = d/2 we nd
ae
i
N
d
=

U
V
e
i/2
Continuity at the left interface gives
ae
i
N
d
=

V
U
e
i/2
As a result,
e
2i(
N
d/2)
=
i
_
[[
2
2
+i
_
[[
2
2
or
e
2i(
N
d/2)
= e
2ii
Since
N
= /hv
x
= /h[v
x
[ we nd
= h[
x
[
_
2
+ arcsin

[[
+
_
l
1
2
__
(3.23)
Combining this with Eq. (3.20) we nally obtain
= h[
x
[
_
2
arcsin

[[
+
_
l
1
2
__
(3.24)
The upper sign refers to k
x
> 0, the lower sign refers to k
x
< 0.
Normalization of the wave function [A[
2
is determined from
_
_
[u[
2
+[v[
2
_
dx = 1
3.2. ANDREEV STATES 61
0 2
@ /2
3@ /2
x
x
0 2
(a) (b)
Figure 3.6: The Andreev bound states in (a) short,
x
[[, and (b) long,
x
[[, SNS structures. The solid lines refer to v
x
> 0, while the dotted lines
are for v
x
< 0.
In the normal region [u[
2
+[v[
2
= 2[A[
2
. In the right region
[u[
2
+[v[
2
= [d
1
[
2
e
2
S
x
_
[
U[
2
+[
V [
2
_
= 2[d
1
[
2
e
2
S
x
[
U[
2
since [
U[
2
= [
V [
2
. Employing the continuity of the wave functions [A[
2
=
[d
1
[
2
e
S
d
[
U[
2
we nd that in the right region
[u[
2
+[v[
2
= 2[A[
2
e
2
S
(xd/2)
In the left region we similarly obtain
[u[
2
+[v[
2
= 2[A[
2
e
2
S
(x+d/2)
Calculating the integrals from to d/2 then from d/2 to d/2 and from
d/2 to we nd
[A[
2
=
1
2(d +
1
S
)
=
1
2
_
[[
2
2
h[v
x
[ +d
_
[[
2
2
(3.25)
Consider the limit of a small width of normal region
d h[v
x
[/[[
such that d cos where
hv
F
/[[
is the coherence length, and is the angle between the momentum k and the x
axis. We have
x
[[. Eq. (3.24) gives
2
arcsin

[[
+
_
l
1
2
_
= 0
0
2
@ /2
3@ /2
x
x
0
2
(a) (b)
@ /2
3@ /2
x
x
Figure 3.7: The Andreev bound states in (a) short,
x
[[, and (b) long,
x
[[, SNS structures for extended energy range [[ < < [[. The solid
lines refer to v
x
> 0, while the dotted lines are for v
x
< 0.
Recall that = arcsin[/[[]. For v
x
> 0 we have to choose l = 1 to have
/2 < < /2. For v
x
< 0, the choice is l = 0. Therefore, the spectrum
becomes [see Fig. 3.6 (a)]
= [[ cos

2
(3.26)
For a long N region, d
0
, when
d h[v
x
[/[[ ,
x
[[
we can neglect arcsin(/[[) so that [see Fig. 3.6 (b)]
= h[
x
[
_
2

2
_
+h[
x
[l (3.27)
where l = 0, 1, . . ..
It is useful here to consider negative energies of excitations. Using Eqs.
(2.41) and (2.42) we notice that if a state with > 0 belongs to k
x
> 0 then the
state with < 0 belongs to k
x
< 0. Therefore, the spectrum Eq. (2.41) now
allows both signs of for any phase dierence in the range 0 < < 2.
For short junctions, the spectrum becomes
= [[ cos(/2) (3.28)
for 0 < < 2 with the upper sign for v
x
> 0 and lower sign for v
x
< 0.
Similarly, for long junctions,
= h[
x
[
_
2

2
_
+h[
x
[l (3.29)
where l = 0, 1, 2, . . . for the entire region 0 < < 2. These spectra are
shown in Fig. 3.7
3.3. SUPERCURRENT THROUGHANSNS STRUCTURE. PROXIMITYEFFECT63
3.3 Supercurrent through an SNS structure. Prox-
imity eect
The current is given by Eq. (2.85)
j =
ihe
m
n
[f
n
(u
n
(r)u
n
(r)) + (1 f
n
) (v
n
(r)v
n
(r)) c.c] (3.30)
where n labels various quantum states. It is more convenient to calculate it
in the N region. Applying the semi-classical approximation, we calculate the
derivatives only of the rapidly varying functions e
ikr
. Using Eqs. (3.17) and
(3.25) and multiplying the current density by the area of the contact S we obtain
for the total current carried by the states with < [[
I
x
=
e
h
n
(1 2f
n
)
hv
x
_
[[
2
2
n
h[v
x
[ +d
_
[[
2
2
n
Here
n
is the bound state energy determined by Eq. (3.24). The quantum
number n describes various states, i.e., the states that belong to various k
y
, k
z
and k
x
() within the area S.
To calculate the supercurrent we assume that there is no voltage across the
junction and the distribution functions correspond to equilibrium.
For the bound states, at a given phase dierence there is only a nite
number of states satisfying the conditions either =
>
(k
x
) or =
<
(k
x
) for
dierent signs of k
x
. Therefore,
I
x
=
e
h
_
_

n,k
x
>0
(1 2f(
>
))
hv
x
_
[[
2
2
>
hv
x
+d
_
[[
2
2
>
n,k
x
<0
(1 2f(
<
))
h[v
x
[
_
[[
2
2
<
h[v
x
[ +d
_
[[
2
2
<
_
_
(3.31)
3.3.1 Short junctions. Point contacts
One can check that for short contacts d , the states in the continuum
> [[ (see Problem 3.3) do not contribute to the current: the contributions
of particles ying from the left and of those ying from the right cancel in
equilibrium. Therefore, Eq. (3.31) gives the full expression for the current.
The simplest case is the limit of short junctions d
0
when one can neglect
the term with d in the energy spectrum. The latter is given by Eq. (3.26). Since
for a given there exists only one state for either v
x
> 0 or v
x
< 0 according
to
>,<
=
where
= [[ cos

2
x
S S
Figure 3.8: The point contact SNS structure.
we nd
I =
eN
>
[[ sin(/2)
h
tanh
[[ cos(/2)
2T
(3.32)
N
>
is the total number of states per unit volume with v
x
> 0 and all possible
k
y
and k
z
ying through the contact of an area S. Eq. (3.32) can be written as
I =
2eN
>
h
tanh

2T
(3.33)
We will see that this is a very general form for a current through superconducting
junctions.
Eq. (3.32) can be applied only when the phase in the superconducting
regions is constant in space, i.e., that it does not vary at distances of the order
of
1
S
, i.e., when the current is smaller than the critical current in the
superconductor. For a wide contact where N
>
in the normal part is of the same
order as in superconductors, it is only true when the phase dierence is small,
1.
In a general case Eq. (3.32) holds if the number of transverse modes in the
normal part N
>
is much smaller than that in the superconducting regions, which
ensures a small value of the current. One of the examples is the so called point
contact, i.e., a structure where two superconductors are connected through a
narrow (with a cross section of an area S a
2
where the transverse dimension
is a ) and short d normal piece called constriction (see Fig. 3.8). In
fact, since the wave function has no possibility to vary within the constriction,
the results do not change if the constriction is also superconducting.
To calculate the current through the point contact it is more convenient to
start with Eq. (3.31). The energy spectrum of a point contact is given by Eq.
(3.26); the normalization of the wave function is
[A[
2
=
_
[[
2
2
2hv
F
(3.34)
(see Problem 3.4). The current through the point contact is found from Eq.
(3.31)
I =
e
h
_
k
x
>0
(1 2f(
>
))
_
[[
2
2
>
k
x
<0
(1 2f(
<
))
_
[[
2
2
<
_
[v
x
[
v
F
=
e[[ sin(/2)
h
tanh
[[ cos(/2)
2T
v
x
>0
v
x
v
F
(3.35)
Counting of the states can be performed as follows. First we write
_
4Sp
2
F
dp
(2h)
3
d
p
4
We put p = p
F
+ h
N
= p
F
+/v
F
and write for a discrete spectrum
dp
2h
= d [(
>
) +(
<
)] (3.36)
Indeed, in our case the integral
_
dp/2h = 2 since there is exactly one state

per unit volume for a given phase dierence for p p
F
.
We note here that while p in Eq. (3.36) can be both p p
F
, the
energy only assumes positive values. As a result,
n
2hN(0)v
F
S
_
>0
d
d
p
4
[(
>
) +(
<
)]
Therefore
v
x
>0
v
x
v
F
= 2hN(0)v
F
S
_
>0
d
_
v
x
>0
d
v
4
v
x
v
F
[(
>
) +(
<
)]
= 2hN(0)v
F
S
_
v
x
>0
d
v
4
v
x
v
F
=
hN(0)v
F
S
2
Finally,
I =
N(0)v
F
Se[[ sin(/2)
2
tanh
[[ cos(/2)
2T
=
[[ sin(/2)
eR
Sh
tanh
[[ cos(/2)
2T
(3.37)
where
1
R
Sh
=
e
2
N(0)v
F
S
2
=
e
2
h
k
2
F
S
(2)
2
is the so called Sharvin conductance (inverse resistance) of the contact in the
normal state. One can write it as
1
R
Sh
=
N
>
R
0
(3.38)
0
2
1
2
3
Figure 3.9: The supercurrent through the point contact. Curve (1) corresponds
to a low temperature T T
c
, curve (3) is for a temperature close to T
c
.
Here
N
>
=
k
2
F
S
(2)
2
is the eective number of states (for both spin projections) penetrating through
the contact. Each state that contributes to the current is called the conducting
channel or conducting transverse mode. The quantity
R
0
=
h
e
2
12.9 k
is the (two-spin) quantum of resistance (see the next Chapter).
The dependence Eq. (3.37) is shown in Fig. 3.9 for various temperatures.
It has a maximum which is called the critical current: A point contact cannot
sustain nondissipative currents larger than I
c
. For low temperatures, the critical
current
I
c
=
[[
eR
Sh
is reached near = . For temperatures close to T
c
the critical current
I
c
=
[[
2
4TeR
Sh
is reached at = /2.
3.3.2 Long junctions
In long junctions, the states with energies larger than [[ do also contribute to
the supercurrent. However, in some cases their contribution is negligible. Here
we consider the current at low temperatures, when the main contribution comes
from the bound states.
For long junctions d
0
we neglect the term hv
x
in the denominator in
Eq. (3.31) and nd
I
x
=
e
d
_
_

n,k
x
>0
[v
x
[ (1 2f(
>
))
n,k
x
<0
[v
x
[ (1 2f(
<
))
_
_
=
e
d
k
x
>0
[v
x
[
_
l
0
l=0
tanh
_
h[
x
[( )/2 + h[
x
[l
2T
_
l
0
l=1
tanh
_
h[
x
[( )/2 + h[
x
[l
2T
_
_
=
e
d
k
x
>0
[v
x
[
l
0
l=l
0
tanh
_
h[
x
[( )/2 + h[
x
[l
2T
_
Here l
0
corresponds to = [[, i.e., l
0
= [[/h[
x
[ 1.
Consider the limit of low temperatures T [[ and very long junction,
[
x
[ T i.e., d hv
F
/T. In the sum over l one can replace the upper limit
by innity. In this case all what happens at energies > [[ has no eect, thus
the delocalized states can be ignored. The sum becomes
=
l=
tanh
_
h[
x
[( )/2 + h[
x
[l
2T
_
If one considers the variable h[
x
[l as continuous and replaces the summation
with integration, the sum turns to zero being a sum of an odd function. There-
fore, we need to take into account the discrete nature of summation. To do this,
we write
l=
tanh
_
h[
x
[( )/2 + h[
x
[l
2T
_
=
1
2i
_
C
1
tanh
_
h[
x
[( )/2 + h[
x
[z
2T
_
cot z dz
where the contour C
1
in the complex plane is shown in Fig. 3.10. It goes around
all poles of cot z at z = l. By shifting the contour into the upper and lower
half-planes we obtain
=
1
2i
_
C
2
tanh
_
h[
x
[( )/2 + h[
x
[z
2T
_
cot z dz
=
T
ih[
x
[
_
C
2
tanh
_
h[
x
[( )/2
2T
+ z
_
cot
_
2T z
h[
x
[
_
d z
=
2T
h[
x
[
n=
cot
_
2T z
n
h[
x
[
_
=
2T
h[
x
[
n=
tanh
n
sin i cos
tanh
n
cos +i sin
Im z
Re z
z
n
l
C
1
C
1
C
2
C
2
Figure 3.10: The contours of integration in the complex plane z.
=
2T
h[
x
[
_

n=0
tanh
n
sin i cos
tanh
n
cos +i sin
+c.c.
_
where
z
n
= i
_
n +
1
2
_
h[
x
[( )/2
2T
are the poles of the tanh function, and
n
=
n
/[
x
[ , = ( )/2
where
h
n
= 2T
_
n +
1
2
_
are the so called Matsubara frequencies.
In the limit of very long junction, [
x
[ T i.e., d hv
F
/T the factor
n
1 so that
tanh
n
1 2e
2
n
Therefore, only the term with n = 0 is important in the sum and
tanh
0
sin i cos
tanh
0
cos +i sin
= i 2ie
2
0
2i
Therefore,
=
8T
h[
x
[
e
2
0
sin =
8Td
h[v
x
[
e
2Td/ h|v
x
|
sin
The current becomes
I
x
=
8Te sin
h
v
x
>0
e
2Td/ hv
x
(3.39)
To sum over the states with v
x
> 0 we write
v
x
>0
S
_
2p
dp
dp
x
(2h)
3
=
2p
2
F
S
(2h)
3
_
/2
0
sin cos d
_
dp
x
=
N(0)v
F
S
2
_
1
0
xdx
_
dp
x
where p
= p
F
sin is the projection of the momentum on the (y, z) plane while
is the angle between the momentum and the x axis, and x = cos . We have
similarly to Eqs. (3.36), (??)
dp
x
= 2h[(
>
) +(
<
)] d
Therefore
I
x
=
4N(0)v
F
STe sin
h
_
1
0
xdx
_
dp
x
e
2Td/ hv
x
= 8N(0)v
F
STe sin
_
1
0
xdxe
2Td/ hv
F
x
We put x = 1/y and nd
_
1
0
xdxe
2Td/ hv
x
=
_

1
dy
y
3
e
(2Td/ hv
F
)y
_

1
e
(2Td/ hv
F
)y
dy =
hv
F
2Td
e
2Td/ hv
F
since hv
F
/2Td 1. Finally,
I
x
=
4hN(0)v
2
F
Se
d
e
d/
N
sin =
1
2eR
Sh
8hv
F
d
e
d/
N
sin (3.40)
where
N
=
hv
F
2T
is the normal-state coherence length. Eq. (3.40) can be written as
I = I
c
sin
where the critical current is
I
c
=
1
2eR
Sh
8hv
F
d
e
d/
N
(3.41)
where 2R
Sh
is the resistance of two SN contacts in the normal state.
We see that for I < I
c
the supercurrent can ow through the normal region
that is in contact with superconductors. This is called the proximity eect: the
superconductor induces Cooper-pair-like correlations between electrons in the
normal state. These correlations decay exponentially into the normal metal over
distances
N
inversely proportional to temperature. The exponential decay of
correlations ensures a small value of the current which is required for the validity
of our calculations.
There is not only the supercurrent in the normal metal but also an energy
gap. Indeed, the energy spectrum is given by Eq. (3.27). The lowest energy of
excitation is
0
= h[
x
[
_
2

2
_
It depends on the phase dierence between the superconductors. The gap van-
ishes for = .
Problems
Problem 3.1.
Find the deection angle of the trajectory during the Andreev reection.
Problem 3.2.
Calculate the velocity of a slow drift of an Andreev state in a SNS structure
with d
0
along the SN plane. Explain the origin of the drift.
Problem 3.3.
Find the wave functions for an SNS structure for > [[.
Problem 3.4.
Find the energy spectrum and wave functions of the superconducting point
contact for < [[
Chapter 4
SuperconductorInsulator
Normal-metal Interface
4.1 Transmission through the barrier
4.1.1 Transmission and reection probabilities
Consider a junction such that the superconductor and the normal-metal half-
spaces are separated by an insulating barrier in the (y, z) plane described by
the potential U(x). We assume that depends only on the x coordinate. The
momentum projections along the y, z plane are conserved. We start from the
BdG equations
h
2
2m
_
ie
hc
A
_
2
u E
F
u + v = u (4.1)
h
2
2m
_
+
ie
hc
A
_
2
v +E
F
v +
u = v (4.2)
Consider the wave function in the form
e
ik
y
y+ik
z
z
_
u(x)
v(x)
_
where k
y
= k
F
sin cos , k
z
= k
F
sin sin , and and are the incidence and
azimuthal angle. The BdG equations are
_
h
2
2m
d
2
dx
2
E
x
+U(x)
_
u + v = u (4.3)
h
2
2m
d
2
dx
2
E
x
+U(x)
_
v +
u = v (4.4)
where
E
x
= E
F

h
2
(k
2
y
+k
2
z
)
2m
=
h
2
k
2
x
2m
73
74CHAPTER 4. SUPERCONDUCTORINSULATORNORMAL-METAL INTERFACE
and k
x
= k
F
cos .
Assume that a particle is incident from the normal region on the left. It will
be reected back by both normal and Andreev reection processes.
We model the insulating barrier by a function potential U(x) = I(x).
The BdG equations take the form
_
h
2
2m
d
2
dx
2
E
x
+I(x)
_
u + v = u (4.5)
h
2
2m
d
2
dx
2
E
x
+I(x)
_
v +
u = v (4.6)
The boundary conditions at the barrier x = 0 are
_
u(0)
v(0)
_
L
=
_
u(0)
v(0)
_
R
(4.7)
and
d
dx
_
u(0)
v(0)
_
R
d
dx
_
u(0)
v(0)
_
L
= 2[k
x
[Z
_
u(0)
v(0)
_
(4.8)
We introduce here the barrier strength
Z =
mI
h
2
[k
x
[
and assume that Z 1. The barrier strength Z is generally a function of the
incident angle , where k
x
= k
F
cos .
Scattering states
Consider rst the case > [[.
There are four independent solutions: (1) The wave function that has an
incident particle on the left of the barrier, (2) the wave function with an incident
e
e
e h
h
I
x 0
N S
Figure 4.1: The NIS structure.
4.1. TRANSMISSION THROUGH THE BARRIER 75
p
F
p
F
-p
F
-p
F

N
0
S
0
i a b c d
Figure 4.2: Left: The state in N region with an incoming particle (i) that is
Andrev reected as a hole (a) and normally reected as a particle (b). Right:
Transmitted particle (c) and hole (d) in S region.
hole on the left, (3) an incident particle on the right, and (4) an incident hole
on the right.
(1) The wave function with an incident particle from the normal region (from
the left) has the form (see Fig. 4.2)
_
u(x)
v(x)
_
L
= e
iq
+
(N)x
_
1
0
_
+ae
iq
(N)x
_
0
1
_
+be
iq
+
(N)x
_
1
0
_
(4.9)
Here
q
(N) = k
x

hv
x
The wave contains normally reected particle with the amplitude b, and Andreev
reected hole with the amplitude a. On the right of the barrier it has transmitted
particle c and a hole d
_
u(x)
v(x)
_
R
= ce
iq
+
(S)x
_
U
V
_
+de
iq
(S)x
_
V
U
_
(4.10)
Here
q
(S) = k
x
2
[[
2
hv
x
and
U =
1
2
_
1 +
_
2
[[
2
_
1/2
, V =
1
2
_
1
_
2
[[
2
_
1/2
The boundary conditions yield
1 +b = cU +dV
a = cV +dU
and
q
(S)V d q
+
(S)Uc q
+
(N)b +q
+
(N) = 2i[k
x
[Z(1 +b)
q
(S)Ud q
+
(S)V c +q
(N)a = 2i[k
x
[Za
We notice that q
+
+ q
2k
x
while q
+
q
/hv
F
so that q
+
q

q
+
+q
and also q
+
q
k
F
Z. As a result, in the semiclassical approximation,
a =
UV
U
2
+ (U
2
V
2
)Z
2
(4.11)
b =
(U
2
V
2
)(Z
2
+iZ)
U
2
+ (U
2
V
2
)Z
2
=
(U
2
V
2
)[Z[
Z
2
+ 1e
i
U
2
+ (U
2
V
2
)Z
2
(4.12)
c =
(1 iZ)U
U
2
+ (U
2
V
2
)Z
2
=
i
Z
2
+ 1Ue
i
U
2
+ (U
2
V
2
)Z
2
(4.13)
d =
iZV
U
2
+ (U
2
V
2
)Z
2
(4.14)
where the scattering phase is dened as
tan = 1/Z
In the limit of the normal state V = 0, U = 1, and a = d = 0 while
b =
iZ
1 +iZ
, c =
1
1 +iZ
so that
[b[
2
=
Z
2
1 +Z
2
, [c[
2
=
1
1 +Z
2
Without a barrier Z = 0 and b = d = 0 while
a =
V
U
, c =
1
U
as it should be for purely Andreev reection, Eq. (3.12).
(2) The state with an incident hole in the normal region is
_
u(x)
v(x)
_
L
= e
iq
(N)x
_
0
1
_
+a
2
e
iq
+
(N)x
_
1
0
_
+b
2
e
iq
(N)x
_
0
1
_
(4.15)
It contains normally reected hole with the amplitude b
2
, and Andreev reected
particle with the amplitude a
2
. On the right of the barrier it has transmitted
hole c
2
and a particle d
2
_
u(x)
v(x)
_
R
= c
2
e
iq
(S)x
_
V
U
_
+d
2
e
iq
+
(S)x
_
U
V
_
The coecients a
2
, b
2
, c
2
, and d
2
are given by Eqs. (4.11)(4.14) where Z is
replaced with Z:
a
2
= a(Z) , b
2
= b(Z) , c
2
= c(Z) , d
2
= d(Z)
p
F
p
F
-p
F
-p
F

N
0
S
0
i a b c d
3 3 3
3
Figure 4.3: The states of incoming, reected and transmitted particles and
holes at a NIS interface when a incident particle approaches barrier from the
superconducting side.
(3) Consider now the process when a particle is incident on a barrier from
the superconducting side, i.e., from the right (see Fig. 4.3).
On the right of the barrier (in the S region) we have
_
u(x)
v(x)
_
R
= e
iq
+
(S)x
_
U
V
_
+a
3
e
iq
(S)x
_
V
U
_
+b
3
e
iq
+
(S)x
_
U
V
_
On the left of the barrier (in the N region)
_
u(x)
v(x)
_
L
= c
3
e
iq
+
(N)x
_
1
0
_
+d
3
e
iq
(N)x
_
0
1
_
The boundary condition require
U +a
3
V +b
3
U = c
3
V +a
3
U +b
3
V = d
3
and
q
+
(N)c
3
q
+
(S)U q
(S)V a
3
+q
+
(S)Ub
3
= 2i[k
x
[Zc
3
q
(N)d
3
q
+
(S)V q
(S)Ua
3
+q
+
(S)V b
3
= 2i[k
x
[Zd
3
As a result, we nd
a
3
= a(Z) , b
3
= b(Z) (4.16)
while
c
3
=
v
g
(S)
v
g
(N)
c(Z) (4.17)
d
3
=
v
g
(S)
v
g
(N)
d(Z) (4.18)
Here
v
g
(S) =
hk
x
m
_
2
[[
2

is the group velocity of particles (holes) in the superconductor. Similarly,
v
g
(N) =
hk
x
m
is the group velocity of particles (holes) in the normal metal.
The state (4) has an incident hole on the right
_
u(x)
v(x)
_
R
= e
iq
(S)x
_
V
U
_
+a
4
e
iq
+
(S)x
_
U
V
_
+b
4
e
iq
(S)x
_
V
U
_
On the left of the barrier (in the N region)
_
u(x)
v(x)
_
L
= c
4
e
iq
(N)x
_
0
1
_
+d
4
e
iq
+
(N)x
_
1
0
_
The coecients a
4
, b
4
, c
4
, and d
4
are given by Eqs. (4.16)(4.18) where Z is
replaced with Z.
The states (2) and (4) can be obtained from (1) and (3), respectively, by
changing signs of and k
x
in q
and in U, V with simultaneous change of 1 to

0 and vice versa (which also corresponds to the change of U into V and vice
versa) in the normal region coherence factors. The reection and transmission
coecients [a[
2
to [d[
2
are determined Eqs. (4.11)(4.14) and (4.16)(4.18).
Thus, to comply with the transformation the factors U and V in the
coecients [a[
2
, [b[
2
, [c[
2
and [d[
2
should be considered as even functions of :
U
2
=
1
2
_
1 +
_
2
[[
2
2
_
, V
2
=
1
2
_
1
_
2
[[
2
2
_
With this denition, the reection and transmission coecients [a[
2
to [d[
2
are
even functions of .
Subgap states
Consider now the case < [[. For the state (1) with k
x
> 0 we have on the
right of the barrier in the superconducting region only the decaying waves
_
u(x)
v(x)
_
R
= ce
i q
+
(S)x
_

U
V
_
+de
i q
(S)x
_

V
U
_
where
q
(S) = k
x
im
h
2
k
x
_
[[
2
2
and
U =
1
2
_
1 +i
_
[[
2
_
1/2
,

V =
1
2
_
1 i
_
[[
2
_
1/2
(4.19)
The expressions Eqs. (4.11)(4.14) for the coecients hold with the replacement
U

U and V

V . However, since

U and

V are complex, we nd
[a[
2
=
[[
2
2
+ ([[
2
2
)(1 + 2Z
2
)
2
(4.20)
and
[a[
2
+[b[
2
= 1 (4.21)
in accordance with the fact that the quasiparticle ow on the right vanishes P
R
=
0. In other words, the incident particle is reected backwards as a combination
of a particle and a hole. Without a barrier, [a[
2
= 1, i.e., the particle is fully
reected as a hole as it should be, Eq. (3.16).
4.1.2 Probability conservation
As we know, the BdG equations (4.1), (4.2) conserve the quasiparticle ow Eq.
(2.90) div P = 0 where the quasiparticle fulx is dened by Eq. (2.89). Consider
the quasiparticle ow in the contact. The ow Eq. (2.89) on the left is according
to Eq. (4.9)
P
L
= 2q
+
(N)
_
1 [b[
2
_
2q
(N)[a[
2
Flow on the right is, see Eq. (4.10)
P
R
= q
+
(S)[c[
2
_
U
2
V
2
_
+q
(S)[d[
2
_
U
2
V
2
_
+c
d [q
(S) (V U UV ) +q
+
(S) (V U UV )] e
iq
(S)xiq
+
(S)x
+c.c
= 2
_
q
+
(S)[c[
2
+q
(S)[d[
2
_ _
U
2
V
2
_
The conservation of ow yields
q
+
(N)
_
1 [b[
2
_
q
(N)[a[
2
=
_
q
+
(S)[c[
2
+q
(S)[d[
2
2
[[
2
or
q
+
(N)[b[
2
+q
(N)[a[
2
+
_
q
+
(S)[c[
2
+q
(S)[d[
2
2
[[
2
= q
+
(N) (4.22)
Note that
hq
(S)
_
2
[[
2
= mv
g,
(S)
where v
g,
(S) is the group velocity of particles (holes) in the superconductor,
which coincide in the quasiclassical approximation. Similarly,
hq
(N) = mv
g,
(N)
where v
g,
(N) is the group velocity of particles (holes) in the normal metal.
Therefore, the ow conservation implies the conservation of quasiparticle cur-
rent probabilities: The sum of current probability of normally reected particle
v
g+
(N)[b[
2
, Andreev reected hole v
g
(N)[a[
2
, transmitted particle v
g+
(S)[c[
2
,
and transmitted hole v
g
(S)[d[
2
is equal to the incident ow v
g+
(N):
v
g+
(N)[b[
2
+v
g
(N)[a[
2
+v
g+
(S)[c[
2
+v
g
(S)[d[
2
= v
g+
(N) (4.23)
The coecients in Eqs. (4.11) (4.14) satisfy Eq. (4.23).
Similarly, equations (4.16)(4.18) and the ow conservation Eq. (4.23) for
the direct process of Fig. 4.2 comply with the conservation of ow in the reverse
process of Fig. 4.3:
v
g
(S)[a
3
[
2
+v
g
(S)[b
3
[
2
+v
g
(N)[c
3
[
2
+v
g
(N)[d
3
[
2
= v
g
(S) (4.24)
Since the group velocities and the DOS are coupled through
N
S
N
N
=
v
F
v
g
(S)
(4.25)
we nd from Eqs.(4.17), (4.18)
(v
F
[c
3
[
2
)N
S
= (v
g
[c[
2
)N
N
, (v
F
[d
3
[
2
)N
S
= (v
g
[d[
2
)N
N
(4.26)
and also
(v
g
(S)[a
3
[
2
)N
S
= (v
F
[a[
2
)N
N
, (v
g
(S)[b
3
[
2
)N
S
= (v
F
[b[
2
)N
N
(4.27)
which express the detailed balance of particle ow within an energy interval dE.
4.2 Current through the NIS junction
Consider the case when a voltage V is applied across the interface. We assume
that the potential of the superconductor is zero while the potential of the normal
metal is V .
The current is
j =
ihe
m
n=1,...,4;,p
_
f
n,,p
_
u
n,,p
u
n,,p
_
+ (1 f
n,,p
)
_
v
n,,p
v
n,,p
_
c.c
(4.28)
It is more convenient to calculate it in the normal region. We have
j =
ihe
m
n=1,3;
_
f
n,
_
u
n,,p
u
n,,p
_
+ (1 f
n,
)
_
v
n,,p
v
n,,p
_
c.c
ihe
m
n=2,4;
_
f
n,
_
u
n,,p
u
n,,p
_
+ (1 f
n,
)
_
v
n,,p
v
n,,p
_
c.c
The state (1) has particles incident from the normal region [see Fig. 4.4(a)].
Therefore, their distribution corresponds to that in the normal region, f
1
=
f
(N). The sum over the states is
n=1
=
_
>0
N
N
d
_
k
x
>0
d
k
4
4.2. CURRENT THROUGH THE NIS JUNCTION 81
p
F
p
F
-p
F
-p
F

N
0
S
0
i a b c d
p
F
p
F
-p
F
-p
F

N
0
S
0
i a b c d
(a)
(b)
3 3 3
3
Figure 4.4: The states of incoming, reected and transmitted particles and holes
at a NIS interface that contribute to the current.
We have
iu
1,,p
x
u
1,,p
+c.c. = 2k
x
_
1 [b[
2
_
, iv
1,,p
x
v
1,,p
+c.c. = 2k
x
[a[
2
The state (2) has holes incident from the normal region [see Fig. 4.4(a)].
Therefore, their distribution corresponds to the distribution of holes in the nor-
mal region, f
2
= 1 f
(N). To get this distribution we note that the hole

excitation with an energy > 0 is the absence of particle in a state below the
Fermi surface, i.e., with an energy . We have
iu
2,,p
x
u
2,,p
+c.c. = 2k
x
[a
2
[
2
, iv
2,,p
x
v
2,,p
+c.c. = 2k
x
(1 [b
2
[
2
)
At the same time, the state (3) has particles incident from the supercon-
ducting region [see Fig. 4.4(b)]. Therefore, their distribution corresponds to
that in the superconductor f
3
= f
(S). The sum over the states is
n=3
=
_
>0
N
S
d
_
k
x
>0
d
k
4
We have
iu
3,,p
x
u
3,,p
+c.c. = 2k
x
[c
3
[
2
, iv
3,,p
x
v
3,,p
+c.c. = 2k
x
[d
3
[
2
Similarly, the state (4) has holes incident from the superconducting region [see
Fig. 4.4(b)]. Therefore, their distribution is f
4
= 1 f
(S), while
iu
4,,p
x
u
4,,p
+c.c. = 2k
x
[d
4
[
2
, iv
4,,p
x
v
4,,p
+c.c. = 2k
x
[c
4
[
2
Therefore, the total current through the junction area S is
I
NIS
=
2heS
m
_
N
N
d
d
4
k
x
_
(1 [b[
2
)(f
(N) +f
(N)) [a[
2
(2 f
(N) f
(N))
2heS
m
_
N
S
d
d
4
k
x
_
_
[c[
2
(f
(S) +f
(S)) +[d[
2
(2 f
(S) f
(S))
v
2
g
(S)
v
2
g
(N)
_
=
2heS
m
_
N
N
d
d
4
k
x
_
(1 [b[
2
+[a[
2
)(f
(N) +f
(N) 1)
([c[
2
[d[
2
)(f
(S) +f
(S) 1)
v
g
(S)
v
g
(N)
_
+
2heS
m
_
N
N
d
d
4
k
x
_
1 [b[
2
[a[
2
([d[
2
+[c[
2
)
v
g
(S)
v
g
(N)
_
We use here
v
g
(S)
v
g
(N)
N
S
= N
N
The last line vanishes due to the conservation of quasiparticle ow. As a result
I
NIS
=
2heS
m
_
>0
N
N
d
_
k
x
>0
d
k
4
k
x
__
1 [b[
2
+[a[
2
[f
(N) +f
(N) 1]
_
[c[
2
[d[
2
[f
(S) +f
(S) 1]
v
g
(S)
v
g
(N)
_
In the normal region, f
(N) = f
0
( eV ) where
f
0
() =
1
e
/T
+ 1
is the equilibrium Fermi function. We thus have
f
(N) +f
(N) 1 = f
0
( eV ) +f
0
( eV ) 1 = f
0
( eV ) f
0
( +eV )
which is an even function of . In the superconductor, f
(S) = f
0
(). Therefore,
f
(S) +f
(S) 1 = 0
The current becomes
I
NIS
=
2heS
m
_
>0
d
_
k
x
>0
d
k
4
k
x
_
1 [b[
2
+[a[
2
[f
(N) +f
(N) 1] N
N
=
heS
m
_

d
_
k
x
>0
d
k
4
k
x
_
1 [b[
2
+[a[
2
[f
(N) +f
(N) 1] N
N
.
eIR
___N
eV __
Z=0
Z=50
Z=1
2
1
0
1 2
I
exc
Figure 4.5: The currentvoltage curves for a NIS interface at low temperatures
[11]. The dashed line is the Ohms law.
We note that
_

B() [f
0
( eV ) f
0
( +eV )] d = 2
_

B() [f
0
( eV ) f
0
()] d
where B() = 1 [b[
2
+[a[
2
is an even function. Finally, if Z is independent of
the incident angle,
I
NIS
= AeN(0)v
F
S
_

d
_
1 [b[
2
+[a[
2
[f
0
( eV ) f
0
()] (4.29)
where A 1 is a constant that depends on the geometry.
The factor
1 [b[
2
+[a[
2
plays the role of the transmission coecient for particle-hole reection at the
NIS interface. It is by the Andreev reection coecient [a[
2
larger that that in
the normal state. The coecients [a[
2
and [b[
2
are given by Eqs. (4.11), (4.12),
(4.20), and (4.21).
The currentvoltage curves for various barrier strengths at low temperatures,
T T
c
are shown in Fig. 4.5.
4.2.1 Normal tunnel resistance
Consider several limiting cases. First assume that the superconductor is in the
normal state [[ = 0. We have [a[
2
= 0 and
[b[
2
=
Z
2
1 +Z
2
The current becomes
I
NIN
=
AeN(0)v
F
S
1 +Z
2
_

[f
0
( eV ) f
0
()] d =
Ae
2
N(0)v
F
SV
1 +Z
2
=
V
R
N
where
1
R
N
=
Ae
2
N(0)v
F
S
1 +Z
2
(4.30)
We use here
_

[f
0
( eV ) f
0
()] d = eV . (4.31)
To prove this we write
_

[f
0
( eV ) f
0
()] d
=
_

[f
0
( eV ) (eV )] d
_

[f
0
() ()] d
+
_

[(eV ) ()] d
Since the rst integral in the r.h.s. converges, we can now make the shift
eV in it, after which it cancels the second integral. The third term
gives
_
eV
0
d = eV
The current through the NIS junction can be written as
I
NIS
=
1 +Z
2
eR
N
_

d
_
1 [b[
2
+[a[
2
[f
0
( eV ) f
0
()] (4.32)
4.2.2 Landauer formula
For Z = 0 we have form Eq. (4.30)
1
R
N
= Ae
2
N(0)v
F
S
1
R
Sh
(4.33)
It is the inverse Sharvin resistance which exists without a barrier.
To explain this result we should recall our assumption that the two conduct-
ing electrodes separated by a contact have dierent voltages. For Z = 0 this
may happen only if the contact has an area much smaller than the cross sections
of the two electrodes. This is exactly similar to the point contacts considered
in the previous chapter.
Consider a point contact between two normal metals in more detail. We
assume that the barrier is absent so that the electrons y freely (ballistically)
through the constriction from one electrode to another. The current through
the constriction is
I = e
p
x
>0;p
y
,p
z
[[v
x
[f
0
( eV ) [v
x
[f
0
()]
= 2e
p
y
,p
z
_

0
dp
x
2h
p
y
,p
z
(p
x
)
p
x
[f
0
( eV ) f
0
()]
The factor 2 comes due to the spin. The summation runs over such p
y
, p
z
whose
states penetrate through the constriction. We further have
I =
2e
h
p
y
,p
z
_

E
F
d [f
0
( eV ) f
0
()] =
2e
2
h
V
p
y
,p
z
_

df
0
()
d
d
=
2e
2
N
>
h
V (4.34)
Here N
>
is the number of states with p
x
> 0 that go through the constriction.
Equation (4.34) is the well-known Landauer formula for a ballistic constric-
tion [12]. It shows that the conductance of the ballistic constriction G =
(2e
2
/h)N is an integer multiple of the quantum of conductance G
0
= 2e
2
/h
where
R
0
=
1
G
0
=
h
2e
2
12.9 k (4.35)
is the quantum of resistance. The dissipation of energy is concentrated in the
electrodes where the incoming particles relax to the local chemical potential.
The Sharvin conductance in Eq. (4.33) can be written as
1
R
Sh
=
2e
2
N
>
h
=
N
>
R
0
where
N
>
= hAN(0)v
F
S = A
2p
2
F
S
(2h)
2
is the number of penetrating modes.
In general, for a contact between two normal metals separated by a constric-
tion with a barrier, the conductance in E. (4.30) can be written as
G
NIN
=
2e
2
h
N
>
n=1
T
n
(4.36)
where
T
n
= 1 [b
n
[
2
=
1
1 +Z
2
n
(4.37)
is the transmission coecient for the mode n. Equation (4.36) is known as the
Landauer-B uttiker formula.
4.2.3 Tunnel current
Consider a junction with a strong barrier Z
2
1. For [[ > [[ we nd from
Eqs. (4.11) and (4.12)
[b[
2
= 1
1
Z
2
(U
2
V
2
)
= 1

Z
2
_
2
[[
2
= 1
N
S
()
Z
2
N(0)
while [a[
2
Z
4
. For [[ < [[ Eqs. (4.20) and (4.21) yield [a[
2
Z
4
thus
[b[
2
= 1 and 1 [b[
2
+[a[
2
= 0.
Therefore,
I
NIS
=
1
eR
N
_

N
S
()
N(0)
[f
0
( eV ) f
0
()] d (4.38)
where we put
N
S
()
N(0)
=

_
2
[[
2
2
[[
2
_
with (x) being the Heaviside step function. This is the well known expres-
sion for the tunnel current. Thus the contact with large barrier strength Z is
equivalent to the tunnel junction.
For low temperatures T T
c
we nd
I
NIS
=
_
(eV )
2
[[
2
eR
N
(eV [[) (4.39)
4.2.4 Excess current
For large voltages, eV [[ the integral in Eq. (4.32) for the current through
the junction is determined by energies of the order of eV . Indeed, for [[
[a[
2
=
[[
2
2
(1 +Z
2
)
2
, [b[
2
=
Z
2
1 +Z
2

[[
2
Z
2
4
2
(1 +Z
2
)
2
Therefore I V/R
N
. The curve I(V ) for large V goes parallel to the Ohms
law, but it is shifted by a constant current which is called the excess current
(see Fig. 4.5). We dene the excess current as
I
exc
(V ) = I
NIS
(V ) I
N
(V )
=
1
eR
N
_

d
_
[1 +Z
2
]
_
1 [b[
2
+[a[
2
1
_
[f
0
( eV ) f
0
()]
The term in the brackets under the integral decays as
2
, therefore the integral
converges at [[ for large voltages and becomes independent of V for eV
[[, T. The current I
exc
thus saturates for high voltages V :
I
exc
() =
1
eR
N
_

d
_
[1 +Z
2
]
_
1 [b[
2
+[a[
2
1
_
[1 f
0
()]
=
1
2eR
N
_

d
_
[1 +Z
2
]
_
1 [b[
2
+[a[
2
1
_
(4.40)
since 1 2f
0
() is an odd function. Here we put f
0
( eV ) = 1 for [[ and
eV [[, T. The current I
exc
() vanishes for Z .
4.3. SIS CONTACT 87
4.2.5 NS Andreev current. Current conversion
One more important limit is for the Andreev-reection mediated current at low
temperatures for a zero barrier strength. For T T
c
and low voltages eV [[
we need only [. We have [a[
2
= 1 while [b[
2
= 0. Therefore, the current
becomes
I
x
=
2
eR
Sh
_

[f
0
( eV ) f
0
()] d =
2V
R
Sh
The conductance is twice the normal-state conductance (see Fig. 4.5). This is
due to the fact that both particles and holes contribute to the current. The
current in the normal region is carried by the normal excitations. However, the
wave function of the normal excitations decays into the superconducting region.
The normal current is then converted into the supercurrent. This follows from
Eq. (2.91)
div J
n
=
2ie
h
(u
n
v
n
u
n
v
n
) ,
according to which the current carried by the subgap states alone is not con-
served in the superconducting region, while the total current is conserved, of
course.
4.3 SIS contact
4.3.1 Wave functions and the energy of bound states
Consider now one more example of a barrier structure which consists of two
superconducting half-spaces separated by a barrier of strength Z in the plane
(y, z). The gap has the form = [[e
/2
in the right (left) superconductor,
respectively. Consider the states with energies < [[.
On the right of the barrier, the decaying wave function has the form
_
u
v
_
= c
2
e
i q
+
x
_

Ue
i/4
V e
i/4
_
+d
2
e
i q
x
_

V e
i/4
Ue
i/4
_
(4.41)
where
q
= k
x
i
_
[[
2
2
h
2
k
x
and
U =
1
2
_
1 +i
_
[[
2
_
1/2
,

V =
1
2
_
1 i
_
[[
2
_
1/2
In the left superconductor,
_
u
v
_
= c
1
e
i q
+
x
_

Ue
i/4
V e
i/4
_
+d
1
e
i q
x
_

V e
i/4
Ue
i/4
_
The boundary conditions Eqs. (4.7), (4.8) yield
c
1

Ue
i/4
+d
1
V e
i/4
= c
2

Ue
i/4
+d
2
V e
i/4
c
1
V e
i/4
+d
1

Ue
i/4
= c
2
V e
i/4
+d
2

Ue
i/4
and
_
c
2

Ue
i/4
d
2
V e
i/4
_
_
c
1

Ue
i/4
+d
1
V e
i/4
_
= 2iZ
_
c
2

Ue
i/4
+d
2
V e
i/4
_
_
c
2
V e
i/4
d
2

Ue
i/4
_
_
c
1
V e
i/4
+d
1

Ue
i/4
_
= 2iZ
_
c
2
V e
i/4
+d
2

Ue
i/4
_
Solving the rst pair of equations for c
1
and d
1
we nd two equations for c
2
and
d
2
c
2

U
__
U
2
+iZ(
U
2

V
2
)
_
e
i/2

V
2
e
i/2
_
+d
2
V
__
V
2
+iZ(
U
2

V
2
)
_
e
i/2

U
2
e
i/2
_
= 0 (4.42)
and
c
2
V
__
V
2
iZ(
U
2

V
2
)
_
e
i/2

U
2
e
i/2
_
+d
2

U
__
U
2
iZ(
U
2

V
2
)
_
e
i/2

V
2
e
i/2
_
= 0 (4.43)
Requiring zero of the determinant we nd the condition of existence of a
nonzero solution
4
U
2
V
2
cos
2
(/2) = 1 +Z
2
_
U
2

V
2
_
2
(4.44)
This yields =
where [13]
= [[
_
1 T sin
2
(/2) (4.45)
where T = 1/(1 + Z
2
) is the transmission coecient in the normal state as
dened by Eq. (4.37). The spectrum of Eq. (4.45) is shown in Fig. 4.6.
Without a barrier Z = 0 when T = 1 we recover the spectrum of a ballistic
point contact Eq. (3.26). For a nal T the gap [[
1 T = [[
appears
for = . Here = 1 T is the reection coecient. The bound state energy
is shifted towards the gap [[ and merges with [[ for a tunnel junction with
very low transmission T 0.
4.3.2 Supercurrent
Using the wave functions Eqs. (4.41) and the spectrum Eq. (4.45) one can
calculate the current through the SIS junction from Eq. (3.30) for an equilibrium
4.3. SIS CONTACT 89
0
2

Figure 4.6: The energy spectrum of a SIS contact
distribution. Near the contact,
s
x 1 we have from Eq. (3.30)
I =
2he
m
n
k
x
Re
_
f(
n
)
_
c
2

U
e
ik
x
x
+d
2

V
e
ik
x
x
__
c
2

Ue
ik
x
x
d
2
V e
ik
x
x
_
[1 f(
n
)]
_
c
2
V e
ik
x
x
+d
2

Ue
ik
x
x
__
c
2

V
e
ik
x
x
d
2

U
e
ik
x
x
__
=
2he
m
n
k
x
[1 2f(
n
)]
_
[c
2
[
2
[d
2
[
2
_
[
U[
2
Using Eqs. (4.42), (4.43) and the normalization of the wave function, we nd
similarly to Eq. (3.32), that the supercurrent is [13] (see Problem 4.4)
I =
N
>
T e[[
2
2h
sin
tanh
_

2T
_
(4.46)
Here N
>
is the number of channels,
N
>
= R
0
/R
Sh
=
h
e
2
R
Sh
It is easy to see that the current can be written in the form of Eq. (3.33)
I =
2eN
>
h
tanh
_

2T
_
where
is given now by Eq. (4.45).

This current is a supercurrent since it ows without voltage. The current
can be written also as
I =
[[
2
2eR
N
sin
tanh
_

2T
_
where
1
R
N
=
T
R
Sh
is the conductance of a contact with a transparency T .
For a ballistic contact T = 1 we recover Eq. (3.37). For a tunnel junction
T 1 the current becomes
I = I
c
sin
where the critical current is [14]
I
c
=
[[
2eR
N
tanh
_
[[
2T
_
(4.47)
4.4 Scattering matrix
Consider the energies [[ < [[. In general, the waves incident on the interface
from the left contain the particle wave e
iq
+
x
with an amplitude u
+
and the
positive group velocity and momentum, and the hole wave e
iq
x
with an
amplitude v
and the negative momentum but positive group velocity. They

are reected at the interface and transformed into reected particle e
iq
+
x
with an amplitude u
(negative momentum and group velocity) and reected

hole e
iq
x
with an amplitude v
+
(positive momentum but negative group
velocity). Reection couples these amplitudes such that
_
u
v
+
_
=

S
_
u
+
v
_
Here
S =
_
S
11
S
12
S
21
S
22
_
is the scattering matrix. The method of scattering matrix for superconducting
tunnel structures was rst used by Beenakker in 1991 [13].
Since for [[ < [[ the ow of quasiparticles Eq. (2.89) to the right from
the interface (in the superconducting region) is zero, it should also vanish in
the normal region due to the conservation of the quasiparticle ow. This is
equivalent to the requirement that

S is a unitary matrix :

S

S = 1 where
=
_
_
t
,
=
_
S
11
S
21
S
12
S
22
_
.
One has
[S
11
[
2
+[S
21
[
2
= 1
[S
22
[
2
+[S
12
[
2
= 1
S
11
S
12
+S
21
S
22
= 0
One can express the components of this matrix through the coecients a
and b calculated previously [see Eqs. (4.11), (4.12]. Consider the wave in the
normal region determined by Eq. (4.9)
_
u(x)
v(x)
_
L
= e
iq
+
(N)x
_
1
0
_
+ae
iq
(N)x
_
0
1
_
+be
iq
+
(N)x
_
1
0
_
(4.48)
4.4. SCATTERING MATRIX 91
The rst term is the incident particle wave with an amplitude u
+
= 1, the
second is the reected hole wave with an amplitude v
+
= a while the third term
is the reected particle wave with an amplitude u
= b. In other words, the

amplitudes of the waves here are related according to
v
+
= au
+
u
= bu
+
v
= 0
The scattering matrix gives for v
= 0 the relations u
= S
11
u
+
and v
+
=
S
21
u
+
. Therefore, S
21
= a and S
11
= b. Let us take now the incident hole wave
Eq. (4.15) with the incident hole amplitude v
= 1. It has
u
= a(Z)v
v
+
= b(Z)v
u
+
= 0
Therefore, S
12
= a(Z) and S
22
= b(Z).
Finally,
S =
_
b(Z) a(Z)
a(Z) b(Z)
_
and
=
_
b
(Z) a
(Z)
a
(Z) b
(Z)
_
If the superconducting gap has a phase = [[e
i
, the coecients are
S =
_
b(Z) a(Z)e
i
a(Z)e
i
b(Z)
_
_
S
N
e
i
S
A
e
i
S
A
e
i
S
N
e
i
_
(4.49)
where we introduce new amplitudes of Andreev S
A
and normal S
N
reections
S
N
=
(
U
2

V
2
)[Z[
Z
2
+ 1
U
2
+ (
U
2

V
2
)Z
2
, S
A
=

U

V
U
2
+ (
U
2

V
2
)Z
2
The unitarity [S
N
[
2
+[S
A
[
2
= 1 follows from [a[
2
+[b[
2
= 1 and (
U
2
)
=

V
2
.
4.4.1 SINIS structures
Consider a SINIS structure where the normal region occupies d/2 < x < d/2.
The superconductor 1 at d/2 < x has a phase
1
while the superconductor 2
at x > d/2 has a phase
2
. Particles (holes) in N region get Andreev reected
as holes (particles) and normally reected as particles (holes) at each interface
and form bound states in N region with a discrete energy spectrum. Using the
scattering matrix approach it is quite simple to nd the eigenstates and the
energy spectrum.
Consider reections at each interface. At the right end of the normal region
x = d/2 we have
_
u
v
+
_
R
=

S
R
_
u
+
v
_
R
(4.50)
Similarly, at the left end of the normal region x = d/2 we have
_
u
+
v
_
L
=

S
L
_
u
v
+
_
L
(4.51)
The scattering matrixes satisfy

S
R
= S(
2
) and

S
L
= S(
1
). The waves at
dierent ends of the normal channel have dierent phase factors
_
u
_
R
= e
i(+
z
)
_
u
_
L
where
= k
x
d , =
d
hv
x
Therefore,
_
u
_
R
= e
i(
z
+)
_
u
_
L
We have at the right end
_
u
v
+
_
R
=

S
R
_
u
+
v
_
R
=

S
R
e
i(
z
+)
_
u
+
v
_
L
=

S
R
e
i(
z
+)

S
L
_
u
v
+
_
L
Therefore
_
1 e
i(
z
+)

S
R
e
i(
z
+)

S
L
_
_
u
v
+
_
L
= 0
The solvability condition requires
det
_
1 e
i(
z
+)

S
R
e
i(
z
+)

S
L
_
= 0 (4.52)
This equation determines the energy spectrum of the bound states.
Symmetric case
Assume Z
R
= Z
L
. Eq. (4.52) yields
4
U
2
V
2
cos
2
(/2) = 1 +
_
U
2

V
2
_
2
A
_
1 e
2i
_
_
V
2
Z
2
(
U
2

V
2
)
_
2
_
1 e
2i
_
_
U
2
+Z
2
(
U
2

V
2
)
_
2
(4.53)
4.4. SCATTERING MATRIX 93
This equation is an analogue to Eq. (4.44) obtained earlier for a SIS structure.
Here the eective barrier strength is
A = 4Z
2
(1 +Z
2
) sin
2
(4.54)
where
=
+
The barrier strength disappears when sin
= 0, i.e., when
= n
which corresponds to a resonant transmission for Z
R
= Z
L
.
SIS contact
The limit of SIS contact is obtained if we assume d = 0 which implies both
= 0 and = 0. In this case the eective barrier strength
A = 4Z
2
= Z
2
2
corresponds to a barrier of a double thickness Z
2
= 2Z. Equation (4.53) becomes
4
U
2
V
2
cos
2
(/2) = 1 +Z
2
2
_
U
2

V
2
_
2
which coincides with Eq. (4.44).
4.4.2 Interference eects in short contacts
Consider now contacts of a nal length such that ,= 0. We consider rst short
contacts such that can be neglected. This requires generally d . We have
from Eq. (4.53)
4
U
2
V
2
cos
2
(/2) = 1 + (
U
2

V
2
)
2
A (4.55)
where the eective barrier strength is given by Eq. (4.54). This gives the
spectrum
2
[[
2
= 1 T sin
2
(/2) (4.56)
of Eq. (4.45) with the transmission coecient
T =
1
1 +A
The barrier strength Eq. (4.54) vanishes A = 0 in the resonance when sin
= 0
and the transmission becomes unity.
Problem 4.1
Derive Eq. (4.39).
Problem 4.2
Calculate the dierential conductance of a tunnel NIS junction at low tem-
peratures and show that it is proportional to the density of states in the super-
conductor at an energy = eV .
Problem 4.3
Calculate the saturated excess current for Z = 0.
Problem 4.4
Derive Eq. (4.46).
Chapter 5
Weak links
5.1 Josephson eect
We have seen that a supercurrent can ow through a junction of two super-
conductors separated by narrow constriction, by a normal region or by a high-
resistance insulating barrier, or by combinations of these. The current is a func-
tion of the phase dierence between the two superconductors. These junctions
are called weak links.
There may be various dependencies of the current on the phase dierence.
The form of this dependence and the maximum supercurrent depend on the
conductance of the junction: The smaller is the conductance the closer is the
dependence to a simple sinusoidal shape. The examples considered in the previ-
ous Chapters are: ballistic contact Eq. (3.37) at temperatures close to T
c
, long
SNS structures, Eq. (3.41), and a tunnel junction, Eq. (4.47).
The presence of a supercurrent is a manifestation of the fundamental prop-
erty of the phase coherence that exists between two superconductors separated
by a weak link; it is called the Josephson eect.
5.1.1 D.C and A.C. Josephson eects
The general features of the Josephson eect can be understood using a very
general example of transitions between two superconductors. Assume that two
superconducting pieces are separated by a thin insulating layer. Electrons can
tunnel through this barrier. Assume also that a voltage V is applied between
the two superconductors.
The wave function of superconducting electrons is a sum
=
(t)
of the states
1
and
2
in superconductor 1 or superconductor 2, respectively.
Each wave function
1
and
2
of an uncoupled superconductor, taken separately,
95
96 CHAPTER 5. WEAK LINKS
2
1
eV
Figure 5.1: The Josephson junction of two superconductors separated by an
insulating barrier.
obeys the Schr odinger equation
ih
t
= E
Here E
( = 1, 2) are the energies of the states in uncoupled superconductors

1 and 2.
When these superconductors are coupled, the wave function satises the
Schr odinger equation
ih
t
=

H
where

H is the total Hamiltonian. This equation determines the variations of
the coecients. If the wave functions
are normalized such that

_

dV =
the coecients obey

ih
C
t
=
[H
]C
(t) .
Here
H
=
_

dV
are the matrix elements. The diagonal elements
H
11
= E
1
+e
1
= E
1
+e
V/2 , H
22
= E
2
+e
2
= E
2
e
V/2
correspond to the energies of the state 1 and 2, respectively. The charge of the
Cooper pair is e
= 2e. The o-diagonal matrix elements describe transitions

between the states 1 and 2
H
12
= H
21
= K .
5.1. JOSEPHSON EFFECT 97
The equation becomes
ih
C
1
t
= eV C
1
(t) KC
2
(t) , (5.1)
ih
C
2
t
= KC
1
(t) eV C
2
(t) . (5.2)
The coecients are normalized such that [C
1
[
2
= N
1
, [C
2
[
2
= N
2
where N
1,2
are the number of superconducting electrons in the respective electrodes. We
put
C
1
=
_
N
1
e
i
1
, C
2
=
_
N
2
e
i
2
.
Inserting this into Eqs. (5.1), (5.2) we obtain, separating the real and imaginary
parts
h
dN
1
dt
= 2K
_
N
1
N
2
sin(
2
1
)
h
dN
2
dt
= 2K
_
N
1
N
2
sin(
2
1
)
and
hN
2
d
2
dt
= eV N
2
+K
_
N
1
N
2
cos(
2
1
)
hN
1
d
1
dt
= eV N
1
+K
_
N
1
N
2
cos(
2
1
)
From the rst two equations we obtain the charge conservation N
1
+ N
2
=
const together with the relation
I
s
= I
c
sin (5.3)
where
I
s
= 2e
dN
2
dt
= 2e
dN
1
dt
is the current owing from the rst into the second electrode, I
c
= 4eK
N
1
N
2
/h
is the critical Josephson current, while =
2
1
is the phase dierence.
To interpret the second pair of equations we note that the overall phase of
the device plays no role. Therefore we can put
2
= /2 while
1
= /2. We
nd after subtracting the two equations
h
t
= 2eV . (5.4)
Equation (5.3) has a familiar form and describes the so called d.c. Josephson
eect: The supercurrent can ow through the insulating layer provided there is
an interaction between the superconducting regions. Equation (5.4) describes
the a.c. Josephson eect: the phase dierence grows with time if there is a
voltage between two superconductors. The d.c. and a.c. Josephson eects
are manifestations of the macroscopic quantum nature of superconductivity.
Various devices which employ these eects can be used for observations and
for practical implementations of the quantum properties of the superconducting
state.
1
2
3
4
I
b
I
Ia
Figure 5.2: A SQUID of two Josephson junctions connected in parallel.
5.1.2 Superconducting Quantum Interference Devices
Equation (5.3) form a basis of SQUIDs. Consider a divice consisting of two
Josephson junctions in parallel connected by bulk superconductors, Fig. 5.2.
Let us integrate v
s
dened by Eq. (1.8) along the contour that goes clockwise
all the way inside the superconductors (dashed line in Fig. 5.2). We have
1
+
2
2e
hc
__
3
1
A dl +
_
2
4
A dl
_
= 0
since v
s
= 0 in the bulk. Neglecting the small sections of the contour between
the points 1 and 2 and between 3 and 4, we nd
b
=
2e
hc
_
A dl =
2
0
(5.5)
where
a
=
2
1
and
b
=
4
3
.
The total current through the device is
I = I
c
sin
a
+I
c
sin
b
= 2I
c
cos
_
0
_
sin
_
0
_
.
The maximum current thus depends on the magnetic ux through the loop
I
max
= 2I
c
cos
_
0
_
. (5.6)
Monitoring the current through the SQUID one can measure the magnetic ux.
5.2 Josephson vortices in extended junctions.
Consider two large superconductors 1 and 2 separated by a thin insulating layer
with a thickness d and placed into the magnetic led. The (x, y) plane is in
5.2. JOSEPHSON VORTICES IN EXTENDED JUNCTIONS. 99
the middle of the insulating layer. The superconductor 1 is at z < d/2, the
superconductor 2 occupies the region z > d/2. The magnetic eld is applied
parallel to the insulating layer along the x axis, Fig. 5.3. We choose A =
(0, A
y
, 0). Therefore,
h
x
=
A
y
z
H
x
y
z
Figure 5.3: A long Josephson junction in a magnetic eld.
Deep in the superconductors we have
y
=
2e
hc
A
y
since v
s
= 0 due to the Meissner screening. The eld decays as
h
x
(z) =
_
He
(zd/2)/
2
z > d/2
He
(z+d/2)/
1
z < d/2
We denote H the eld in between the two superconductors at z = 0;
1,2
is
the London penetration length in the superconductor 1 or 2, respectively. The
phase is essentially independent of z because the current through the junction
is small. In the bulk it is thus the same as at the boundary of the contact.
The vector potential deep in the superconductor 2 is
A
y
(2) =
_

0
h
x
dz =
_
d/2
0
h
x
dz
_

d/2
h
x
dz = H
d
2
H
2
Here we put A
y
(0) = 0 and assume that H is independent of z at 0 < z < d/2.
Similarly,
A
y
(1) =
_

0
h
x
dz =
_
0
d/2
h
x
dz +
_
d/2
h
x
dz = H
_
d
2
+
1
_
As a result
H =
hc
2e(
1
+d/2)
1
y
=
hc
2e(
2
+d/2)
2
y
so that
y
=
2e(
1
+
2
+d)H
hc
(5.7)
where
=
2
1
The magnetic eld H is generally a function of the coordinate y into the junction.
Using the Maxwell equation
c
4
curl
z
h = j
c
sin or
c
4
H
y
= j
c
sin
we obtain
hc
2
8e(
2
+
1
+d)
y
2
= j
c
sin
or
2
J
y
2
= sin (5.8)
where
J
=
hc
2
8ej
c
(
2
+
1
+d)
(5.9)
is called the Josephson length. Equation (5.8) is called Ferrell and Prange [1963]
equation.
We can write Eq. (5.7) in the form
H =
4j
c
2
J
c
y
(5.10)
Equation (5.8) is similar to equation of motion for a pendulum. Indeed, the
latter has the form
d
2
dt
2
=
g
l
sin
where the angle is measured from the bottom. Equation (5.8) is obtained
from it by replacing = and putting
2
J
= g/l. This means that the
pendulum angle is measured from the top, Fig. 5.4.
5.2.1 Low eld limit
Consider rst a low eld H applied outside the junction at y = 0. In this case
is also small. Expanding Eq. (5.8) we nd
2
J
y
2
=
whence (see Fig. 5.5, curve 1)
=
0
e
y/
J
5.2. JOSEPHSON VORTICES IN EXTENDED JUNCTIONS. 101
Figure 5.4: A pendulum moves in time similar to variations of along the y

axis.
The magnetic eld decays as
H(y) = H(0)e
y/
J
where we nd from Eq. (5.10)
H(0) =
4j
c
0
c
(5.11)
Magnetic eld decays into the junction in a way similar to the Meissner eect.
The penetration length is
J
. This length is larger than
L
because the screening
current cannot exceed the Josephson critical current j
c
. Such behavior exists
for elds smaller than 8j
c
J
/c.
2
Y
0
2Y
0
y
1
2
3Y
0
Figure 5.5: The phase dierence as a function of the distance into the junc-
tion measured from the left edge. Curve 1: small magnetic elds. Curve 2:
large elds, phase runs from = at the edge through 2n values making
Josephson vortices. The curves correspond to H(0) < 0.
5.2.2 Higher elds. Josephson vortices.
To nd a solution of Eq. (5.8) for larger elds, we multiply it by /y and
obtain
2
J
2
_
y
_
2
+ cos = A (5.12)
where A is a constant. Eq. (5.12) gives
2
_

0
d
Acos
= y (5.13)
As discussed above, the stable solution for y corresponds to = 0 and
/y = 0. Eq. (5.12) results in A = 1. The applied led is then
H(0) =
4j
c
2
J
c
(0)
y
=
8j
c
J
c
sin

0
2
The sign is chosen to agree with Eq. (5.11) for small elds. Increasing eld
leads to an increase in
0
until it reaches . This threshold corresponds to
the led
H
1
=
8j
c
J
c
=

0
J
(
1
+
2
+d)
Above this eld, the constant A > 1, and the phase can vary within
unlimited range. Consider for example the case H(0) < 0. The phase runs
indenitely from at the edge through the values 2n producing the so called
solitons (Fig. 5.5, curve 2). The phase solitons are also called the Josephson
vortices: the phase dierence across the junction varies by 2 each time as we
go past one Josephson vortex. The distance between vortices is L = 2Y
0

J
.
Consider the case H H
1
. In this case A 1 and the slope coincides with
the magnetic eld
y
=
Hc
4j
c
2
J
whence
=
Hcy
4j
c
2
J
+
0
The supercurrent becomes
j
s
= j
c
sin
_
2y
L
+
0
_
where
L =
8
2
j
c
2
J
cH
=

0
(
1
+
2
+d)H
is the distance between the Josephson vortices.
The total current through the junction of the width W is
I = j
c
_
W/2
W/2
sin
_
2y
L
+
0
_
dy = Wj
c
sin(W/L)
W/L
sin
0
Its maximum is
I
max
= I
c
sin(/
0
)
/
0
where I
c
= j
c
W and
= HW(
1
+
2
+d)
is the total ux through the junction.
5.3. DYNAMICS OF JOSEPHSON JUNCTIONS 103
R J V
I
Figure 5.6: The resistively shunted Josephson junction.
5.3 Dynamics of Josephson junctions
5.3.1 Resistively shunted Josephson junction
Here we consider the a.c. Josephson eects in systems which carry both Joseph-
son and normal currents in presence of a voltage. As we know, the normal cur-
rent has a complicated dependence on the applied voltage which is determined
by particular properties of the junction. In this Section, we consider a simple
model that treats the normal current as being produced by usual Ohmic resis-
tance subject to a voltage V . This current should be added to the supercurrent.
Therefore, the total current has the form
I =
V
R
+I
c
sin (5.14)
where the phase dierence is =
2
1
. Since the Josephson current through
the junction is small, the current density in the bulk electrodes is also small.
Thus, the phases
1
and
2
do not vary in the bulk,
1,2
= const. The dierence
of the phases at the both sides from the hole obeys the Josephson relation
h
t
= 2eV (5.15)
This equation describes the so called resistively shunted Josephson junction
(RSJ) model (see Fig. 5.6).
The full equation for the current is
I =
h
2eR
t
+I
c
sin (5.16)
If I I
c
, the phase starts to grow with time, and a voltage appears. Let t
0
be the time needed for the phase to grow from /2 to /2 + 2. The average
voltage is then
(2e/h)V = 2/t
0

J
(5.17)
I I
V
c
Figure 5.7: The currentvoltage curve for resistively shunted Josephson junc-
tion.
Calculating t
0
(see Problem 5.1) we nd the currentvoltage curve dependence
V = R
_
I
2
I
2
c
(5.18)
It is shown in Fig. 5.7.
5.3.2 The role of capacitance
The Josephson junction has also a nite capacitance. Let us discuss its eect
on the dynamic properties of the junction.
The current through the capacitor (see Fig. 5.8) is
I = C
V
t
The total current becomes
I =
hC
2e
t
2
+
h
2eR
t
+I
c
sin (5.19)
Let us discuss this equation. Consider rst the work
A =
_
t
0
IV dt =
h
2e
I
produced by an external current source. We nd
A =
_
t
0
t
_
h
2
C
8e
2
_
t
_
2
hI
c
2e
cos
_
dt +
h
2
4e
2
R
_
t
0
_
t
_
2
dt
This equation has the form of a balance of energy
[E
capacitor
+E
junction
] = A
h
2
4e
2
R
_
t
0
_
t
_
2
dt
where the energy of the capacitor is
E
capacitor
=
h
2
C
8e
2
_
t
_
2
=
CV
2
2
(5.20)
R J C
I
Figure 5.8: The capacitively and resistively shunted Josephson junction.
The energy of the Josephson junction is
E
junction
= E
J
[1 cos ] , where E
J
=
hI
c
2e
(5.21)
The last term in the energy balance is the dissipative function.
Eq. (5.19) can also be written as a mechanical analogue equation
J
t
2
+
t
+E
J
sin = F (5.22)
of a pendulum with the moment of inertia (or mass)
J =
h
2
C
4e
2
=
h
2
8E
C
and the maximum gravity force torque mgl = E
J
in a viscous medium with a
viscosity
=
h
2
4e
2
R
=
h
2
8E
C
RC
under action of a constant torque
F =
hI
2e
Here we introduce the energy
E
C
=
e
2
2C
associated with charging the capacitor C with one electron charge. We will
meet this quantity later when we discuss the Coulomb blockade eects in small
junctions.
The resonance frequency of the pendulum is
p
=
_
E
J
J
=
_
2eI
c
hC
=
_
2cI
c
0
C
=
8E
J
E
C
h
(5.23)
It is called the plasma frequency.
Equation (5.22) can be considered as an equation of motion of a particle
with a coordinate , a mass J in a potential
U() = E
J
[1 cos ] ( hI/2e) = E
J
[1 cos I/I
c
] (5.24)
in presence of viscosity. The potential Eq. (5.24) is called a tilted washboard
potential, Fig. 5.9.
E
Figure 5.9: The tilted washboard potential. The tilting angle is determined
by the ratio I/I
c
. The dot shows a particle with a coordinate in a potential
minimum.
Sometimes it is convenient to introduce an eective inductance equivalent
to the Josephson junction if the phase variations are small. For example, for
small the Josephson current becomes I
J
= I
c
. On the other hand, due to
the Josephson relation,
=
2e
h
_
V dt
Therefore, the Josephson current is
I
J
=
2eI
c
h
_
V dt
It looks like a current through an inductance where the voltage across the in-
ductance is
V =
1
c
t
=
L
c
2
I
t
(in Gaussian units) whence
I =
c
2
L
_
V dt
Therefore, the eective inductance is
L
J
=
hc
2
2eI
c
(5.25)
In terms of the eective inductance, the plasma frequency is
p
=
_
2eI
c
hC
=
c
L
J
C
which coincides with the resonance frequency of an LC circuit.
Equation (5.19) can be written also as
2
p
t
2
+Q
1
1
p
t
+ sin =
I
I
c
(5.26)
where we introduce the quality factor
Q =
p
RC =
_
2eI
c
R
2
C
h
(5.27)
that characterizes the relative dissipation in the system. This parameter is large
when resistance is large so that the normal current and dissipation are small.
Consider the dynamics of the Josephson junction in an increasing current.
As long as the current is below I
c
, the phase is stationary: it is determined
by I = I
c
sin . The junction is superconducting. In the representation of a
mechanical particle with a coordinate in a tilted washboard potential this
means that the particle is localized in one of the minima of the potential (state
0
in Fig. 5.10). As I increases and approaches I
c
, the tilt increases, and the
minima gradually disappear as shown in Fig. 5.10. For I > I
c
the particle
begins to roll down the potential relief. A nonzero velocity /t determines
the voltage across the junction.
E
1
2
0
Figure 5.10: The tilted washboard potential for I/I
c
close to unity. Dashed line:
I I
c
, the minima disappear.
The currentvoltage dependence is most simple for an overdamped junction
which corresponds to small Q i.e., to small capacitance and large dissipation. In
this case we can neglect the term with the second derivative in Eq. (5.22). We
thus return to the case considered in the previous section where the current
voltage dependence is determined by Eq. (5.18).
For a nite Q the currentvoltage dependence becomes hysteretic (see Fig.
5.11). With increasing current voltage is zero and the phase is localized
(state
0
in Fig. 5.10) until I reaches I
c
. For I > I
c
the particle rolls down the
potential (solid line in Fig. 5.10), and a nite voltage appears which corresponds
to a voltage jump shown by a solid line in Fig. 5.11. However, when the current
is decreased, a dissipative regime with a nite voltage extends down to currents
smaller than I
c
. The current at which the voltage disappears is called retrapping
current. It corresponds to trapping of the particle back into one of the potential
minima
0
in Fig. 5.10.
This behavior has a simple explanation. A particle with a small dissipation
will roll down the potential overcoming the potential maxima by inertia even
if I < I
c
provided the loss of energy during its motion from one maximum
(state 1 in Fig. 5.10) to the next (state 2) is smaller than the energy gain
( h/2e)I = hI/e. If the dissipation is larger (i.e., Q is smaller), the energy
loss exceeds the energy gain and the particle has no energy to continue its
motion, thus it falls down into the potential minimum and remains trapped there
(state
0
in Fig. 5.10). In a sense, this describes a transition from insulating
to superconducting state with increasing dissipation.
I I
V
c
I
r
Figure 5.11: The currentvoltage curve for resistively and capacitively shunted
Josephson junction. The dotted line (coinciding with Fig. 5.7) is for resistively
shunted junction, small Q. The solid lines show the hysteretic behavior of a
contact with a large Q.
It can be shown that, for a highly underdamped junction , i.e., for large
Q , the retrapping current goes to zero while the IV curve has the linear
Ohmic dependence. For large Q, the voltage is almost constant V

V , even at
I I
c
and the phase has the form
= 2e
V t/h +
where 1. Indeed, Eq. (5.26) yields for the time-independent component
2e
V
h
=
Q
p
I
I
c
=
2eIR
h
where we use Eqs. (5.23) and (5.27) so that the I-V curve is linear
V = IR
For the oscillating component we have
2
p
t
2
+Q
1
1
p
t
+ sin(
J
t) = 0
where we put
J
=
2e
V
h
For a large Q we neglect the rst derivative and nd
=

2
p
2
J
sin(
J
t)
The variation is small if
p
/
J
1. This condition reads
2
p
J
=
I
c
I
p
RC
=
I
c
IQ
1
Therefore it should be I
c
/IQ 1. If the current does not satisfy this condition,
becomes large, and the nite voltage regime breaks down. Therefore, the
retrapping current is
I
r
I
c
/Q (5.28)
It goes to zero as Q .
5.3.3 Thermal uctuations
Consider rst overdamped junction. A particle with a coordinate is mostly
sitting in one of the minima of the washboard potential in Fig. 5.9. It can
go into the state in a neighboring minimum if it receives the energy enough to
overcome the barrier. This energy can come from the heat bath, for example,
from phonons. The probability of such a process is proportional to exp(U/T)
where U is the height of the barrier as seen from the current state of the particle.
The probability P
+
to jump over the barrier from the state
0
to the state
0
+2
and the probability P
to jump over the barrier from the state

0
+2 back to
the state
0
are
P
=
a
exp
_
U
0
(hI/2e)
T
_
where
a
is a constant attempt frequency, and U
0
is the average barrier height.
Therefore, the probability that the particle will go from the state
0
to the state
0
+ 2 is
P = P
+
P
= 2
a
exp
_
U
0
T
_
sinh
_
hI
2eT
_
This will produce a nite voltage
V =
_
h
2e
_
2P =
2h
a
e
exp
_
U
0
T
_
sinh
_
hI
2eT
_
For low currents, I I
c
, the barrier height is independent of the current
U
0
= 2E
J
. For I 0 we nd
V =

2
h
2
a
I
e
2
T
exp
_
2E
J
T
_
I I
V
c
Figure 5.12: The currentvoltage curves of a RSJ junction in presence of thermal
uctuations. The curves from bottom to top correspond to decreasing E
J
/T;
the curve starting at I = I
c
refers to E
J
= .
This is a linear dependence characterized by certain resistance that depends on
the attempt frequency. One can express the attempt frequency in terms of the
resistance in the normal state R. Indeed, for T
E
J
the Josephson barrier is
ineective thus the exponent can be replaced by unity, and the current voltage
dependence denes the normal resistance
2
h
2
a
e
2
T
= R
whence
a
= e
2
E
J
R/
2
h
2
. Using this we nd for the voltage
V =
E
J
RI
T
exp
_
2E
J
T
_
This determines the eective resistance of the junction [15]
R
J
=
E
J
R
T
exp
_
2E
J
T
_
(5.29)
It is exponentially small for low temperatures.
We see that the junction has a nite (though small) resistance even for
low currents. The currentvoltage curve for an overdamped RSJ junction in
presence of thermal uctuations is shown in Fig. 5.12.
In the case of underdamped junctions, the particle will roll down the po-
tential relief as soon as it gets above the potential barrier. The probability of
this process is just P =
a
exp(U/T). The attempt frequency
a
is now the
oscillation frequency in the potential minimum determined by sin = I/I
c
such
that
2
a
=
2
p
2
cos =
2
p
_
1
I
2
I
2
c
_
1/2
The barrier height is U = U
max
U
min
where U
min
is the value of the energy
Eq. (5.24) at = arcsin(I/I
c
) while U
max
is its value at = arcsin(I/I
c
).
Therefore
U = 2E
J
_
cos arcsin
I
I
c
I
I
c
arccos
I
I
c
_
= 2E
J
1
I
2
I
2
c
2IE
J
I
c
arccos
_
I
I
c
_
(5.30)
The probability is more important for large currents I I
c
when the barrier is
small,
U
4
2
3
E
J
(1 I/I
c
)
3/2
(5.31)
This can be approximated in the whole range as
U 2E
J
(1 I/I
c
)
3/2
As the current increases from zero to I
c
the probability P =
a
exp(U/T)
of an escape from the potential minimum increases from exponentially small
up to P
p
10
10
sec
1
. The voltage generated by escape processes is
V (h/e)P. The measuring device will detect a voltage at a current I
cf
when
it is above the sensitivity limit V
0
. Thus
V
0
=
h
p
e
exp
_
U(I
cf
)
T
_
The threshold current in Fig. 5.11 is thus
I
cf
= I
c
_
1 [(T/2E
J
) ln(h
p
/eV
0
)]
2/3
_
which is smaller than I
c
.
The rising part of the IV curve in Fig. 5.12 for an overdamped junction
near I
c
is also determined by an exponential dependence V = (h/e)P
+
where
the probability P
+
contains the barrier from Eq. (5.31). Indeed, the probability
of the reverse process P
is now strongly suppressed by a considerably higher

barrier seen from the next potential minimum.
5.3.4 Shapiro steps
When a Josephson junction is driven by an a.c. voltage (or is subject to a
microwave irradiation) with a frequency , the d.c. component of supercurrent
through the junction exhibits the so called Shapiro steps: jumps of the current
at constant voltages satisfying V
n
= nh/2e.
Let the voltage across the junction be
V = V
0
+V
1
cos(t)
The phase dierence across the junction is then
=
0
+
J
t + (2eV
1
/h) sin(t)
I
V
I
1
h/2e
2h/2e
3h/2e
Figure 5.13: The currentvoltage curves of a RSJ junction irradiated by a mi-
crowave with frequency .
where
J
= 2eV
0
/h. The supercurrent becomes
I = I
c
sin = I
c
k=
(1)
k
J
k
(2eV
1
/h) sin(
0
+
J
t kt)
where k runs over integer numbers. We use here the expansion
e
iz sin
= J
0
(z) + 2
k=1
J
2k
(z) cos(2k) + 2i
k=0
J
2k+1
(z) sin[(2k + 1)]
=
k=
J
k
(z) cos(k) +i
k=
J
k
(z) sin(k)
with z = 2eV
1
/h and = t. We note that due the parity J
k
(z) = (1)
k
J
k
(z)
of the Bessel functions, the components with odd k drop out from the rst sum
in the second line, while the components with even k drop out from the second
sum. Using this we arrive at the above expression for the current.
We see that for
J
= k, i.e., for
V
k
= kh/2e
the supercurrent has a d.c. component I
k
= I
c
J
k
(2eV
1
/h) sin(
0
+ k). This
d.c. component adds to the total d.c. current and produces the step parallel to
the current axis with the width I
k
= 2I
c
J
k
(2eV
1
/h).
Problems
Problem 5.1.
Find the distance between Josephson vortices for H close to H
1
.
Problem 5.2.
Derive Eq. (5.18).
Problem 5.3
Two Josephson junctions have the critical currents I
c1
= 500 A and I
c2
=
700 A are connected in parallel by superconductors. The total current through
both of them is I = 1 mA. Find the currents through each junction.
Problem 5.4.
The junction has a critical current I
c
= 1 mA and the normal resistance
R = 2 . Find the d.c. voltage and the Josephson frequency
J
if the current
through the junction is I = 1.2 mA.
Problem 5.5.
The critical current of the junction is I
c
. The current through the junction
has d.c. and a.c. components such that
I = I
0
+I
1
sin(t)
where I
0
< I
c
and I
1
I
0
. Find the voltage across the junction.
Chapter 6
Coulomb blockade in
normal double junctions
6.1 Orthodox description of the Coulomb block-
ade
For more detailed description including the eects of environment see, for ex-
ample, review [18].
C
1
, R
1
C
2
, R
2
V
1
V
2
V
g
C
g
Island
Figure 6.1: The equivalent circuit of a SET. The island is coupled to the voltage
source via two contacts with resistances R
1
, R
2
and capacitances C
1
, C
2
, and
to the gate through the capacitor C
g
. The bias voltage is V = V
1
V
2
.
Consider the device called the single electron transistor (SET) with the
equivalent circuit shown in Fig. 6.1. For simplicity we assume a symmetric
situation C
1
= C
2
, R
1
= R
2
R
T
such that V
1
= V/2, V
2
= V/2, and that
the capacitance of the gate C
g
is small. Let the charge on the island provided
by the gate voltage be Q
0
= V
g
C
g
.
For zero bias voltage V = 0, the electrostatic energy of the island having
a charge Q consisting of the continuous oset charge Q
0
provided by the gate
electrode and a discrete charge of k electrons that have tunneled into the island,
115
116CHAPTER 6. COULOMB BLOCKADE INNORMAL DOUBLE JUNCTIONS
e/2 3e/2 -3e/2 0
Q
0
E
e -e -e/2 2e -2e
k=0 k=-1 k=1
Figure 6.2:
Q = ke +Q
0
, is
Q
2
2C
=
(Q
0
+ke)
2
2C
=
Q
2
0
2C
+
ke(Q
0
+ke/2)
C
Here C
= C
1
+C
2
is the total capacitance. The spectrum is shown in Fig. 6.2.
The parabolas intersect at Q
0
= ke/2.
If the temperature is low, T E
C
, the tunneling into the island at small
bias voltage becomes possible and the current can ow through the junction
only for those gate charges for which the parabolas intersect. For other gate
charges, the low-voltage current is zero. Let us consider the conditions for the
current as functions of the bias voltage and the gate charge.
The energy dierence between the state of the island after k electrons have
tunneled from the source which has the bias potential V
b
is
E =
(Q
0
+ke)
2
2C
Q
2
0
2C
keV
b
=
ke(Q
0
+ke/2)
C
keV
b
The dierence vanishes when V
b
= V
b,k
, where
V
b,k
=
Q
0
+ke/2
C
For tunneling of one electron the voltage when the tunneling starts is
V
b,1
=
Q
0
+e/2
C
[see Fig. 6.3 (a)]. It vanishes if the oset charge on the island provided by the
gate is Q
0
= e/2. For this charge, the I-V curve starts from zero voltage, Fig.
6.3 (b).
If the bias voltage is increased, the two-electron tunneling becomes possible
when
V
b,2
=
Q
0
+e
C
6.1. ORTHODOX DESCRIPTION OF THE COULOMB BLOCKADE 117

V
b1 V
b2
V
b3
V
b
I
(a)
V
b2
V
b3 V
b
I
(b)
Figure 6.3: The Coulomb staircase: The current exhibits steps at V = V
b,k
. (a)
Zero oset charge Q0 = 0, when V
b,k
= (ke/2C; (b) Oset charge Q
0
= e.
The steps with k ,= 1 have smaller amplitudes due to smaller probabilities of
tunneling.
and so on. The appearance of the two-electron process is seen on the I-V curve
as another step. The steps associated with multiple-charge tunneling are called
the Coulomb staircase.
Let us consider now the one-electron processes and calculate the tunneling
rates.
In the presence of the bias voltage, the electrostatic energy change in a
state with a charge Q on the island for adding an electron to the normal island
through the left junction is
E
+
L
=
(Q+e)
2
2C
_
Q
2
2C
+
eV
2
_
=
e(Q+e/2)
C
eV
2
= 2E
C
_
n +
1
2
_
eV
2
(6.1)
Here E
C
= e
2
/2C
is the characteristic charging energy of the island, and

n = Q/e = n +Q
0
/e where n is an integer number of extra electrons.
In general, the electrostatic energy change in a state with a charge n =
n + Q
0
/e on the island for adding (+) or removing () an electron to the
normal island through the left junction is
E
L
(n) = 2E
C
( n 1/2) eV/2
The electrostatic energy change in a state with a charge n on the island for
adding (+) or removing () an electron to the normal island through the right
junction is
E
R
(n) = 2E
C
( n 1/2) eV/2
The tunnelling rates are
L(R)
(n) =
1
e
2
R
T
_

dE f
1
(E)[1 f
2
(E E
L(R)
(n))]. (6.2)
Here R
T
is the resistance of one contact, f
1
(E) is the distributions on the source
electrode, while f
2
(E E) is the distribution on the target electrode before
the tunneling event; therefore 1f
2
(EE) is the probability to nd that the
state is empty where the tunneling should occur. The probability of tunneling
is proportional to the transparency of the contact T 1/R
T
. The other factors
in the coecient in front of the integral in are chosen in such a way as to
provide the correct expression for the resistance of the contact in the Ohmic
regime, see Eq. (6.7) below.
For equilibrium distribution f
i
(E) = (1 +e
E/T
i
)
1
with T
1
= T
2
,we have
f(E)[1 f(E E
(n))] =
f(E) f(E E
)
1 exp(E
/T)
and
_

dE [f(E) f(E +x)] = x

which is nothing but Eq. (4.31) proven earlier.
Therefore, Eq. (6.2) yields
(n) =
1
e
2
R
T
E
exp(E
/T) 1
(6.3)
The current into the island through the left (right) junction is
I
L(R)
= e
n=
(n)[
+
L(R)
(n)
L(R)
(n)] (6.4)
where (n) is the probability of having n extra electrons on the island. We have
n=
n(n) = 0 (6.5)
by symmetry, and
n=
(n) = 1 ,
n=
n(n) = Q
0
/e (6.6)
6.1.1 Low temperature limit
For T 0 the tunneling rates Eq. (6.3) are
L(R)
(n) =
1
e
2
R
T
[E
L(R)
[ (E
L(R)
)
The rates vanish when all E
L(R)
are positive. This takes place when
E
(+)
L
> 0 : eV/2 2E
C
( n + 1/2) < 0 (a)
E
()
L
> 0 : eV/2 2E
C
( n 1/2) > 0 (b)
E
(+)
R
> 0 : eV/2 + 2E
C
( n + 1/2) > 0 (c)
E
()
R
> 0 : eV/2 + 2E
C
( n 1/2) < 0 (d)
In Fig. 6.4 the diamond-shaped region in the plane (V, Q = e n) is shaded where
all the rates are zero. This is the region where the current to and out of the
island is zero, and the charge on the island does not change due to the Coulomb
blockade. If the oset charge Q
0
is in the range e/2 < Q
0
< e/2, the state
with n = 0 is stable.
(a)
(b)
(c)
(d)
V
Q e/2 -e/2
e/C
-e/C
Figure 6.4: The region of stable charge is shaded. The lines (a)(d) correspond
to equalities in Eqs. (a) to (d).
V
Q
0
e/2
-e/2 3e/2 -3e/2
n=0 n=1
n=-1
Figure 6.5: The regions of stable charge n = 0, 1, . . . as functions of the gate
charge.
With change in the gate voltage, the stable charge on the island will vary by
integer number of electrons due to tunneling to or from the respective electrode.
The regions of stable states with n = 0, e, . . . as functions of the gate charge
are shown in Fig. 6.5.
6.1.2 Conductance in the high temperature limit
We have for the left junction
+
L
(n)
L
(n) =
1
e
2
R
T
_
E
+
L
exp(E
+
L
/T) 1

E
L
exp(E
L
/T) 1
_
Up to the rst order in E
C
/k
B
T
e
we have
+
L
(n)
L
(n) =
T
e
2
R
T
_
v
_
1
1 e
v
+
1
1 e
v
_
2 nE
C
T
[f(v) +f(v)] +
E
C
T
[f(v) f(v)]
_
=
T
e
2
R
T
_
v
2 nE
C
T
+
E
C
T
[f(v) f(v)]
_
Here we introduce the reduced voltage v eV/2T and denote
f(v) =
1
1 e
v
+
ve
v
(1 e
v
)
2
We also use
f(v) +f(v) =
1
1 e
v
+
1
1 e
v
= 1
Using Eqs. (6.5) and (6.6) we nd
I
L
=
T
eR
T
_
v
2Q
0
E
C
eT
+
E
C
T
[f(v) f(v)]
_
=
T
eR
T
_
v
2Q
0
E
C
eT

E
C
T
sinh v v
2 sinh
2
v
_
We nd for the dierential conductance [19]
G
G
T
= 2R
T
dI
dV
= 1
E
C
T
v sinh v 4 sinh
2
(v/2)
4 sinh
4
(v/2)
(6.7)
where G
1
T
= 2R
T
is the total resistance of the contacts. This equations shows
in particular that the resistance of the junction in the absence of charging eects
is 2R
T
, i.e., the resistance of each tunnel contact is R
T
. This conrms the choice
of the coecient in Eq. (6.2). The depth of the conductance minimum at V = 0
in Eq. (6.7) is
G/G
T
=
E
C
3T
The behavior of the conductance is shown in Fig. 6.6.
V
0
G/G
T
1
T
1
T
2
Figure 6.6: The minimum in conductance as a function of the bias voltage due to
Coulomb eects at hight temperatures. The two temperatures satisfy T
2
< T
1
.
Problem 6.1.
Derive the Ohms law from Eq. (6.4) for E
C
= 0.
Problem 6.2.
Find the dierential conductance of overheated double junction when the
lead temperature is zero while the island temperature is T.
Chapter 7
Quantum phenomena in
Josephson junctions
7.1 Quantization
7.1.1 Quantum conditions
Quantum eects can be observed in Josephson structures consisting, for exam-
ple, of a very small superconducting grain connected to superconducting charge
reservoirs through small tunnel junctions having very low capacitance and high
tunnel resistance, as shown in Fig. 7.1. The equivalent circuit is shown in Fig.
6.1. The necessary constraints can be easily estimated from general arguments.
First, the Coulomb charging energy for one electron e
2
/2C should be larger than
temperature T to avoid thermal smearing of the charge states on the supercon-
ducting island. For T 1 K this gives C < 10
15
F which strongly restricts
the size of the junction by an area 10
8
cm
2
. Second, the tunnel resistance
should be large enough to avoid averaging out by quantum uctuations in the
particle number. To be observable, the charging energy e
2
/2C must exceed the
quantum uncertainty in energy h/t h/RC associated with the nite life-
time of the charge on the capacitor. Equating e
2
/2C to h/RC we nd that
the capacitance drops out and the condition becomes R > R
0
where R
0
is the
resistance quantum R
0
= h/2e
2
12 k, the quantity already familiar from
Eq. (4.35).
Another realization may be a small Josephson junction with a capacitance
C and tunnel resistance R satisfying the above conditions, which is connected
through small-capacitance, C
ext
C, high-resistance contacts such that R
0

R
ext
R to the external current source, Fig. 7.2.
123
124CHAPTER 7. QUANTUMPHENOMENA IN JOSEPHSON JUNCTIONS
V
R,C R,C
L I
G
1
L
2
C
g
Figure 7.1: Realization of the quantum Josephson junction device: A small
island I is connected to the external leads L
1
and L
2
by tunnel contacts. The
tunnel resistance R should be larger than R
0
. An additional gate electrode G
is connected to the island through a capacitor C
g
to control the voltage on the
island.
R
R C
S S
ext
ext
C
V
L
1
L
2
Figure 7.2: Another realization of the quantum Josephson junction device: A
small Josephson junction is connected to the external leads L
1
and L
2
by high-
resistance R
ext
and low capacity C
ext
C contacts. Both R
ext
and the tunnel
resistance R should be larger than R
0
.
7.1.2 Charge operator
Consider an isolated Josephson junction. The charging energy of the capacitor
is
Q
2
2C
=
CV
2
2
=
C
2
_
h
2e
t
_
2
If the phase dierence is treated as a particle coordinate, the time derivative
/t should be considered as a velocity, while the charging energy is equivalent
to the kinetic energy.
In quantum mechanical description, the kinetic energy is written in terms of
the momentum operator. If the coordinate is then the momentum operator
is dened as
p
= ih

(7.1)
This denition complies with the usual commutation rule
[ p
, ]
= ih (7.2)
To nd out the physical meaning of the momentum operator let us consider
the continuity equation for the supercurrent
d(eN
s
)
dt
=
_
j
s x
x
d
3
r
7.1. QUANTIZATION 125
The current density j has the form of the charge ow density en
s
v
s
which is
e
E
s
p
s
where p
s
is the momentum of a superconducting particle and E
s
is the super-
conducting energy density. The momentum of the Cooper pair is 2p
s
= h/x
so that the continuity equation takes the form
d(eN
s
)
dt
= e
_

x
E
s
p
s
d
3
r =
2e
h
_

x
E
s
(/x)
d
3
r =
2e
h
c
s
since the phase gradient over the length x is /x = /x where is the (given)
phase dierence. The quantity
c
s
=
_
E
s
d
3
r
is the superconducting energy. Therefore,
h
2
dN
s
dt
=
c
s
(7.3)
This equation can be considered as one of the Hamiltonian equations p/t =
H/x. Since is the coordinate and c is the energy, the quantity hN
s
/2 is
the momentum of the particle. However, according to Eq. (7.2) the canonically
conjugated momentum operator is p
. Therefore, p
= hN
s
/2 and
i

=
N
s
2
= N
p
(7.4)
is the operator of the number of Cooper pairs N
p
= N
s
/2. The second Hamil-
tonian equation has the form
x
t
=
H
p

t
=
c
( hN
s
/2)
or
1
2
t
=
c
s
( hN
s
)
=

s
h
where
s
is the chemical potential of Cooper pairs. This equation coincides with
the Josephson relation since
s
= eV .
Equation (7.4) denes the operator of superconducting charge transferred
through the junction
Q = e

N
s
= 2e

N
p
= 2ie

(7.5)
The commutation relation takes the form
[

Q, ]
= 2ie
Therefore, the quantum uncertainty in phase and in charge Qare restricted
by the charge of a Cooper pair Q 2e.
The eigenfunction of a state with the charge Q obeys the equation
Q
Q
= Q
Q
or 2ie
= Q
Q
It is
Q
() = Ce
iQ/2e
(7.6)
Assuming a single-valued wave function
Q
( + 2) =
Q
()
we obtain quantization of charge of a Cooper pair Q/e = 2n, i.e.,
Q = 2en
where n is a integer.
Note that is the phase of the wave function of a Cooper pair of electrons.
The single electron phase would be
1
= /2. If we now require a single-valued
one-electron wave function,
Q
(
1
) = Ce
iQ
1
/e
so that (
1
+ 2) = (
1
), we obtain 2Q/e = 2n such that the single-
electron charge is integer: Q = en.
7.1.3 The Hamiltonian
The charging energy of a capacitor can be written as
Q
2
2C
=
4e
2
2C
2
= 4E
C
2
where
E
C
=
e
2
2C
is the the charging energy for the charge of one electron.
The total energy of the Josephson junction becomes
H = 4E
C
2
+E
J
[1 cos ] (7.7)
This is the Hamiltonian of the junction in the quantum mechanical description.
If the junction is connected to a current source, the charge operator changes
Q = 2ie

+q(t)
where q(t) is a continuous charge supplied by the current source. The Hamilto-
nian becomes
H = 4E
C
_
i

+
q(t)
2e
_
2
+E
J
[1 cos ] (7.8)
7.2. MACROSCOPIC QUANTUM TUNNELLING 127
In the classical limit this Hamiltonian is equivalent to the washboard poten-
tial. Indeed, the classical analogue of Eq. (7.8) is
c =
[Q+q(t)]
2
2C
+E
J
[1 cos ]
where Q is the charge that is transferred through the junction. Using the
Josephson relation V = ( h/2e)(d/dt) the charging energy can be transformed
as
[Q+q(t)]
2
2C
=
Q
2
2C
+
[Q+q(t)]q(t)
C

q
2
(t)
2C
=
Q
2
2C
+V q(t)
q
2
(t)
2C
=
Q
2
2C

h
2e
dq
dt
+
d
dt
_
h
2e
q(t)
_
q
2
(t)
2C
=
Q
2
2C

h
2e
I +
dF(t)
dt
Here I = dq/dt. The last term dF(t)/dt is a full derivative of certain function.
It can be omitted. With Eq. (7.5) for the operator Q, the total Hamiltonian
assumes the usual form of a Hamiltonian of a particle in the tilted washboard
potential
H = 4E
C
2
+E
J
[1 cos ]
hI
2e
(7.9)
The quantum-mechanical description goes over into the classical picture de-
scribed in Chapter 5 when the charge Q in the charge eigen-function Eq. (7.6) is
large as compared to the electron charge and can be considered as a continuous
variable.
7.2 Macroscopic quantum tunnelling
With the account of quantum eects, the behavior of the junction in presence
of a high bias current is dierent from that considered in the previous chapter.
Consider the Hamiltonian Eq. (7.9) for a representative particle in a washboard
potential. The representative particle with the coordinate can now escape
from the potential minimum at
0
by tunnelling through the potential barrier,
see Fig. 7.3, maximum 1. If the maximum 2 in Fig. 7.3 is lower than the
minimum
0
, the particle needs one tunnelling through the barrier shown by a
gray region in the gure.
Tunnelling of the representative particle means a tunnelling of the entire
system from one macroscopic state that contains many particles to another
macroscopic state. This process involves a macroscopic number of particles
and thus its probability should be inherently small. However, the Josephson
junction provides a tool that can help us to observe these macroscopic quantum
tunnelling (MQT) events.
The easiest way to solve the Schrodinger equation
_
4E
C
2
+U()
_
() = E()
E
1
2
0
0
Figure 7.3: The tilted washboard potential in the quantum case. A quantum
particle can escape from the potential minimum by tunnelling through the bar-
rier (grey region).
with the washboard potential
U() = E
J
_
1 cos
I
I
c
_
is to use the WKB approximation
= exp
_
i
_
() d
_
assuming d/d
2
. We nd
2
=
E U
J
()
4E
C
The WKB approximation holds if dU/d
3
E
C
or when E
J
E
C
.
For the energy below the potential maximum we have
= i
=
i
_
U
J
() E
2E
1/2
C
which ensures the decay of the wave function for positive . The transmission
probability through the barrier is proportional to the square of the transmission
amplitude
exp
_
d
_
where
o
and
are the turning points satisfying E = U(). The probability of

tunnelling becomes
P
a
exp
_
E
1/2
C
_

0
_
U
J
() E d
_
(7.10)
The exponent is generally of the order of
(E
J
/E
C
)
1/2
1
7.2. MACROSCOPIC QUANTUM TUNNELLING 129
where =
0
. This results in a very small probability. For zero current,
. Writing h
p
= (8E
J
E
C
)
1/2
we can present the probability as
P
p
exp (2E
B
/h
p
)
where E
B
2E
J
is the barrier height. This will transform into the Boltzmann
factor exp(E
B
/T) for the crossover temperature
T
cr
h
p
/2
For typical value of
p
10
11
sec
1
this corresponds to T
cr
100 mK.
The tunnelling probability increases for I I
c
, when the barrier height is
getting small, see Eq. (5.31). We have
U
4
2
3
E
J
(1 I/I
c
)
3/2
while
= arccos(I/I
c
) =
_
1 (I
2
/I
2
c
)
so that the factor in the exponent for the probability becomes
(E
J
/E
C
)
1/2
[1 (I/I
c
)]
5/4
7.2.1 Eects of dissipation on MQT
For low temperatures, the system occupies the low energy states in the potential
minimum with the oscillator frequency
p
. Consider the limit of low currents.
The characteristic time it takes for the system to tunnel through the barrier
is t
t
2/
p
. The energy dissipated during this time is
E
D

h
2
4e
2
R
_
d
dt
_
2
t
t

2h
2
p
4e
2
R
It should be smaller than the energy itself, E
D
h
p
/2, otherwise the sys-
tem cannot tunnel into a state in another potential minimum. This gives the
condition
R R
0
=
h
2e
2
R
0
being the quantum of resistance. If this condition is fullled, the MQT is
possible. The phase can escape from the potential minimum, and the current
driven junction will exhibit a nite voltage. It will not be superconducting in a
strict sense. However, if the dissipation is larger, i.e., R < R
0
, the phase cannot
tunnel. There will be no voltage: the junction is superconducting. Therefore,
the dissipation helps the superconductivity, which is a counterintuitive result.
We can look at this estimate also in a dierent way. When the phase is xed
to one of the potential minima, the charge Q on the superconducting island
is not dened due to the quantum uncertainty relation. Thus, the quantum
uctuations of charge are large. On the contrary, when the phase can tunnel,
its uncertainty increases and the charge becomes more localized. This agrees
with the estimates on the barrier resistance made in the beginning of the present
Chapter.
7.3 Band structure
7.3.1 Blochs theorem
The Band structure of the energy states in a periodic potential is a consequence
of the Blochs theorem known in solid state physics [16]: Any solution of the
Schr odinger equation for a particle in a potential U(x) periodic with a period a
has the form
k
(x) = u
k
(x)e
ikx
where u
k
(x) is a periodic function
u
k
(x +a) = u
k
(x)
Am equivalent formulation of the Blochs theorem is that for a particle in a
potential U(x) periodic with a period a there exists a quantity k such that the
wave function obeys
k
(x +a) = e
ika
k
(x) (7.11)
The quantity k is called quasimomentum. The energy, i.e., the eigenvalue of the
Schr odinger equation
_
h
2
2m
d
2
dx
2
+U(x)
_
k
(x) = E
k
k
(x)
depends on the quasimomentum. The energy spectrum is split into intervals
continuously lled by the values E
k
as functions of k (energy bands) separated
by intervals where there no values of E
k
(forbidden bands). These energy bands
are labelled by the band numbers n such that E = E
kn
.
The quasimomentum is dened within an interval
a
k

a
which is called the rst Brillouin zone. All the quasimomenta that dier by an
integer multiple of 2/a are equivalent, i.e., the quasimomenta
k
= k + (2/a)n
refer to the same quasimomentum. Indeed, Eq. (7.11) shows that
k
(x +a) =
e
ika
k
(x), i.e., belongs to the same quasimomentum as
k
. However, there
may be many states belonging to the same quasimomentum, so that k and
k + (2/a)n do not necessarily belong to the same state. An example can be
constructed for a free particle with a spectrum E = p
2
/2m if one introduces
a very small (zero) potential with an (arbitrary) period a. This spectrum is
shown in Fig. 7.4.
Since the quasimomenta /a and /a dier by 2/a, the points at the right
and left boundary of the Brillouin zone are equivalent. One can thus consider
the so called extended zone scheme where the energy is periodic as a function
of quasimomentum with a period 2/a. This is shown in Fig. 7.4 by dashed
curves.
7.3. BAND STRUCTURE 131
/a /a 3/a 3/a 0
k
E
Figure 7.4: The energy spectrum for a free particle in the presence of a small
periodic potential. The spectrum shown by solid lines is reduced to the rst
Brillouin zone /a < k < /a. The dashed lines refer to the extended zone
scheme.
7.3.2 Blochs theorem in Josephson devices
In the case of Josephson junctions, the coordinate is . If the junction is not
connected to the current source, the period of the Josephson potential is 2.
Therefore, solutions of the Schr odinger equation with the Hamiltonian Eq. (7.7)
4E
C
k
+E
J
[1 cos ]
k
= E
k
(7.12)
should obey
k
( + 2) = e
i2k
k
() (7.13)
where k is dened within the rst Brillouin zone 1/2 < k < 1/2. Equation
(7.12) is known in mathematics as the Mathieu equation.
Without the potential we would have
k
= e
ik
Comparing this with Eq. (7.6) we recognize that k plays the role of charge Q/2e.
Therefore, the quasimometum k in the presence of potential is the quasicharge
Q = 2ek
dened within the rst Brillouin zone
e < Q < e (7.14)
If we require a single-valued wave function
k
(+2) =
k
() we nd that
k = n so that the quasicharge dened within the rst Brillouin zone is zero, i.e.,
Q = 0. The energies of a free charge (see Fig. 7.4) are
E
Q
= 4E
C
n
2
=
(2en)
2
2C
which corresponds to an integer number of electron pairs on the junction.
The situation changes if we have an external current source. The Hamilto-
nian has the form of Eq. (7.8). The Schr odinger equation becomes
4E
C
_
i

+
q(t)
2e
_
2
+E
J
[1 cos ] = E (7.15)
We make a gauge transformation
=

e
iq(t)/2e
where the function

satises Eq. (7.12) and obeys the Blochs theorem, Eq.
(7.13), i.e.,
Q
( + 2) = e
iQ/e
Q
() (7.16)
As a result the function satises
Q
( + 2) = e
i[Qq(t)]/e
Q
() (7.17)
Requiring it to be single valued we nd
Q = q(t) + 2en (7.18)
whence
dQ
dt
=
dq
dt
= I (7.19)
We see that the quantum mechanics of the Josephson junction connected to
the current source can be described by the Hamiltonian Eq. (7.7) where the
quasimomentum depends on time according to Eq. (7.19).
7.3.3 Large Coulomb energy: Free-phase limit
This limit is realized when the Josephson energy E
J
is much smaller that the
charging energy E
C
, i.e., E
J
E
C
. The Schr odinger equation (7.12)
4E
C
Q
+E
J
[1 cos ]
Q
= E
Q
Q
(7.20)
It has the solutions which are close to the eigenstates for xed charge Eq. (7.6).
The spectrum has the form of parabolas
E E
J
= E
C
Q
2
e
2
=
(q + 2en)
2
2C
shown in Fig. 7.5. The parabolas are shifted by integer multiple of the Cooper
pair charge 2e.
The quantum-mechanical description implies that the the charge q in Fig. 7.5
is replaced by a quasicharge Q reduced to the rst Brillouin zone, e < Q < e.
7.3. BAND STRUCTURE 133
e 3e 3e 0
q
E
2e 2e e 4e 4e
Figure 7.5: The energy spectrum of a free charge (in a zero Josephson potential)
as a function of the bias charge q.
e e 3e 3e 0 Q
E
E
g
Figure 7.6: The energy spectrum for a Josephson junction in the limit of nearly
free phase. The spectrum in the rst Brillouin zone e < Q < e is shown by
solid lines.
A small Josephson potential introduces small energy gaps at the boundary
of the Brillouin zone where the free-charge parabolas cross (black point in Fig.
7.5). To calculate the rst energy gap we note that the potential
E
J
cos =
E
J
2
_
e
i
+e
i
couples the states at the Q = e boundary of the Brillouin zone
Q=e
= e
i/2
and the states at the Q = e boundary
Q=e
= e
i/2
which dier by Q = 2e and thus belong to the same quasicharge Q. The wave
function at Q = e will thus be a linear combination
e
= c
1
e
i/2
+c
2
e
i/2
Inserting this into the Schr odinger equation (7.20) we nd
E
C
_
c
1
e
i/2
+c
2
e
i/2
_
E
J
2
_
c
1
e
3i/2
+c
2
e
i/2
+c
1
e
i/2
+c
2
e
3i/2
_
= (E E
J
)
_
c
1
e
i/2
+c
2
e
i/2
_
The harmonics with 3i/2 couple to the Q = 3e quasicharge. Comparing the
coecients at the i/2 harmonics we nd
(E
C
E +E
J
)c
1
E
J
2
c
2
= 0
(E
C
E +E
J
)c
2
E
J
2
c
1
= 0
whence
E = (E
C
+E
J
)
E
J
2
This means that the energy gap has the width E
g
= E
J
with the middle at
E
C
+ E
J
, see Fig. 7.6. The middle point is shifted with respect to its free-
phase-limit (E
J
= 0) location at E
C
due to the constant component of the
potential. The lowest energy is also shifted above zero, see Problem 7.1.
Since the boundaries Q = e and Q = e of the Brillouin zone are equivalent,
as well as they are, in general, for any Q = 2em (m is an integer), one can use the
so called extended zone scheme where the energy in each band E
Q,n
E
n
(Q)
is a periodic function of Q:
E
n
(Q+ 2em) = E
n
(Q)
This is shown by dashed lines in Figs. 7.4, 7.6.
7.3.4 Low Coulomb energy: Tight binding limit
In this limit the Josephson energy is larger than the charging energy E
J
E
C
which implies large capacitance. The system behavior is close to that for a
particle in a series of deep potential wells. One can expand the potential near
each minimum
U() =
E
J
2
2
to get the oscillator potential. The Schr odinger equation (7.20) transforms into
the oscillator equation
4E
C
d
2
d
2
+
E
J
2
2
= E
The energy spectrum is
E
n
=
_
8E
C
E
J
_
n +
1
2
_
= h
p
_
n +
1
2
_
7.4. COULOMB BLOCKADE 135
U
E
b
0 2 2
Figure 7.7: The energy band spectrum for a Josephson junction in the limit of
large Josephson energy (tight binding). The energy bands are widened oscillator
levels for a localized particle with the wave function .
The energy spacing h
p
E
J
_
E
C
/E
J
E
J
. The lowest energy wave function
is
0
() = C exp
_
2
4
_
E
J
2E
C
_
Due to the periodic nature of the potential one can construct the true wave
function
Q,n
() =
m
e
i2m(Q/2e)
n
( 2m)
where
n
( 2m) is a function centered at = 2m. The function
n
() is
called the Wannier function [16]. It is close to the wave function
n
() obtained
by solving the equation near each minimum. This wave function
Q,n
() sat-
ises the Bloch condition Eq. (7.16). Each level is broadened into an energy
band (see Problem 7.3)
E
Q
= E
n
1
2
E
b,n
cos
Q
e
(7.21)
The band width is determined by overlaps of the wave functions
n
() centered
at
m
= 2m. It is exponentially small. For example, the lowest band width is
E
b,0
= 32
_
E
J
E
C
_
1/2
_
E
J
2E
C
_
1/2
exp
_
_
8E
J
E
C
_
The quantum properties of Josephson junctions are discussed in review [17].
7.4 Coulomb blockade
Let us consider Fig. 7.5. If we increase, by a bias current from the external
source, the charge starting from q = 0 for n = 0 the energy will increase until
it reaches the crossing point (black dot) at q = e corresponding to the charge
e on the capacitor and to the voltage e/C across the capacitor. With a further
increase in q the system will go over to a state with n = 1 corresponding to
the parabola shifted by 2e to the right that has a lower energy. The transition
from n = 0 to n = 1 corresponds to the 2e charge transfer through the
tunnel Josephson junction. We see that the charge transfer (current) through
the capacitor occurs only when the voltage reaches a threshold value V
C
= e/C.
This is the Coulomb blockade: preventing of the charge transfer by the charging
energy. To describe the Coulomb blockade quantum-mechanically, we need rst
to consider the semi-classical equation of motion in a periodic potential.
7.4.1 Equation of motion
In the semi-classical theory than neglects transition between energy bands, the
equation of motion for the quasimomentum of a particle is [16]
h
k
t
= F
For a constant force F this gives hk = Ft. The velocity is
x
t
=
E
n
hk
where E
n
(k) is the band energy. This yields
x
hk
=
1
F
E
n
hk
or
x = F
1
E
n
(k) , hk = Ft
For a free particle in zero potential E(k) = h
2
k
2
/2m, so that the particle is
continuously accelerated. However, in a periodic potential, E(k) is a periodic
function. Therefore, instead of being accelerated, the coordinate of particle
performs Bloch oscillations with the amplitude x = 2E
b
/F and a period
t =
2h
aF
In a rst Brillouin zone picture, the particle is moving until it is reected
at the zone boundary such that its quasimomentum changes from k
B
= /a at
one boundary to k
B
2/a = k
B
at another boundary.
7.4.2 Bloch oscillations and the Coulomb blockade in Joseph-
son junctions
Consider low currents such that the (Zener) transitions from one band to another
have low probability. In a Josephson junction, the force equation has the form
Q
t
= I (7.22)
while
t
=
E
Q,n
hk
=
2e
h
E
Q,n
Q
(7.23)
e/C
-e/C
V
t
Figure 7.8: The voltage across the junction as a function of time for a constant
current bias.
Therefore, for a constant current,
Q
=
2e
Ih
E
Q,n
Q
so that
=
2e
Ih
E
n
(Q)
The period of Bloch oscillations is from Eq. (7.22)
t
B
=
2e
I
The amplitude of Bloch oscillations is
=
2e
Ih
E
b
In the limit of large Coulomb energy (free phase), E
C
E
J
E
b
E
C
and
E
C
/E
J
t
B
/R
0
C 1
The phase is not xed: it oscillates rapidly with a large amplitude. The voltage
is
V =
h
2e
t
=
E
Q,n
Q

Q
C
Each time when the quasicharge Q approaches the boundary of the Brillouin
zone Q
B
= +e (or Q
B
= e), the quasicharge changes Q
B
Q
B
2e such that
the voltage jumps from +e/C to e/C (or vice versa). The average voltage is
zero.
The change in the quasicharge Q by 2e means the 2e charge transfer through
the Josephson junction. We see that the charge transfer through the junction
occurs only when the voltage across the junction reaches a threshold value e/C.
This is the quantum-mechanical picture of the Coulomb blockade. The charging
energy of the junction prevents the charge transfer through it unless the voltage
U
E
b
0 2 2
Figure 7.9: MQT is equivalent to LandauZener transitions between the energy
bands up to the continuum. The energy bands are widened oscillator levels for
a localized particle with the wave function .
exceeds the threshold. At the threshold V
C
= e/C, the charging energy Q
2
/2C
of a charge Q = e becomes equal to the charging energy (Q 2e)
2
/2C of the
charge Q = e 2e = e on the capacitor after the Cooper pair has tunnelled
through the junction. To see the Coulomb blockade one needs a junction with
a rather low capacity.
On the contrary, if the capacity is high such that E
J
E
C
, the band width
is very narrow, and the amplitude of phase oscillations is exponentially small.
The phase is essentially xed such that the current I
c
cos ows without voltage:
the junction is superconducting. A nite voltage can then appear as a result of
macroscopic quantum tunnelling considered in the previous Section within the
semiclassical approach. In the semiclassical picture of Eqs. (7.22) and (7.23),
the macroscopic quantum tunnelling is equivalent to Zener transitions from a
lower band up to higher bands in Fig. 7.7 and nally to the continuum for
E > E
J
(see Fig. 7.9). Neglecting the Zener transitions implies absence of
MQT and assumes that the bias current is small.
7.4.3 Eect of dissipation
R J C
I
Figure 7.10: The resistively shunted Josephson junction.
Consider the resistively shunted junction, Fig. 7.10, described by [17]
I =
Q
n
t
+
V
R
(7.24)
where, as before, the voltage across the junction
V =
E
Q,n
Q
V/V
th
I/I
th
1
1
Figure 7.11: The d.c. I-V curve for resistively shunted quantum junction for
the limit of high Coulomb energy. V
th
= I
th
R = e/C.
We again neglect the inter-band transitions assuming that the current is smaller
that what is required for Zener transitions.
For high Coulomb energy,
E
Q,n
Q
=
Q
C
Eq. (7.24) becomes
Q
n
t
= I
Q
RC
(7.25)
Assume that the charge is within the rst Brillouin zone e Q +e.
If the current is below the threshold value I I
th
, the Bloch oscillations begin. Eq. (7.25) has the solution
Q = Ae
t/RC
+IRC
For t = 0 we have Q = 0 and Q/t = I. This gives
A = IRC
Assume that at t = t
1
the charge was at the end Q = e of the Brillouin zone,
while at t = t
2
the charge was at the end Q = +e of the Brillouin zone. We nd
Aexp(t
2
/RC) = e IRC , Aexp(t
1
/RC) = e IRC
Therefore
exp
_
t
1
+t
2
RC
_
=
IRC +e
IRC e
The average voltage is found from
(t
1
+t
2
)
V
R
=
_
t
2
t
1
_
A
RC
exp
_
t
RC
_
+I
_
dt
= (t
1
+t
2
)
_
I +
A
t
1
+t
2
[exp(t
1
/RC) exp(t
2
/RC)]
_
Finally,
V
R
= I 2I
th
_
ln
I +I
th
I I
th
_
1
The second term is zero for I I
th
and diverges for I . The I-V curve is
shown in Fig. 7.11. The maximum d.c. voltage V is V
th
= I
th
R = e/C. For
R the I-V curve is vertical which means zero d.c. voltage.
7.5 Parity eects
e/2 e/2 3e/2 3e/2 0
q
E
e 2e 2e e
Figure 7.12: The e-periodic energy dependence of a normal junction as a func-
tion of the continuous charge q supplied by the external source.
Let us consider a junction in Fig. 7.1 made of normal conductors and study
how its properties change when a superconducting gap is introduced.
In the case of a normal junction, the wave function depends on a single-
electron phase
1
= /2 such that the charge operator in Eq. (7.5) becomes
Q = 2ie

= ie

1
(7.26)
The eigenfunction has the form
Q
= e
iQ
1
/e
7.5. PARITY EFFECTS 141
which is 2-periodic for Q = ne. The Josephson current disappears, and the
Hamiltonian becomes
H = E
C
2
1
If the junction is connected to the external leads, the Hamiltonian takes the
form
H = E
C
_
i

1
+
q(t)
e
_
2
The solution of the corresponding Schr odinger equation has the form
= e
i(Qq)
1
/e
The condition of 2-periodicity gives Q = q +me where m is an integer.
0
q
E
e 2e 2e
e
0
q
E
e 2e 2e
e
0
q
E
e 2e 2e
e
E
C
E
C
(a)
(b)
(c)
Figure 7.13: The energy of junctions. (a) e-periodic dependence in the normal
state. (b) Energy of odd-number states is shifted by < E
C
. The charge
transfer occurs at the crossing points (black dots). (c) > E
C
, the charge
transfer occurs with a 2e periodicity.
The energy is
E
Q
= E
C
Q
2
e
2
= E
C
(q +me)
2
e
2
It is shown in Fig. 7.12. The dierent parabolas correspond to dierent values
of m.
Assume we start with m = 0. As we increase the charge q supplied by the
external source, the energy grows until q reaches e/2 which corresponds to the
voltage V
C
= e/2C. At this moment, the energy of the capacitor becomes equal
to the energy for the state with m = 1, i.e., to (q e)
2
/2C. At this point the
charge at the capacitor decreases by e, one electron being transferred through
the junction via tunnelling. We see again that the charge transfer (current) does
not occur unless voltage reaches the threshold value V
C
(Coulomb blockade).
Consider again the device shown in Fig. 7.1 and apply a gate voltage V
G
between the island and the gate electrode. If V
G
= V
C
, the energy of the
junction corresponds to the level where the parabolas cross (black dots in Fig.
7.12). This means that an innitely small voltage V can lead to a continuous
transfer of charge from lead L
1
to lead L
2
.
Let us now assume that the island I is superconducting with an energy gap
. For simplicity we restrict ourselves to zero temperatures. If the number
of electrons on it is even, they all are included into Cooper pairs and form
the ground state with zero energy plus the charging energy. If now we add an
extra (odd) particle, it can only occupy a state above the gap thus the energy
will be plus the charging energy. One more particle will make a pair with
the previous one thus decreasing the total energy down to simply the charging
energy. Therefore, the states with odd number of particles will have energy
which is the charging energy for the given number of particles shifted by +
with respect to the energy for the even number of particles. This is shown in
Fig. 7.13.
Shift of the parabolas for odd particle numbers destroys the e-periodicity
of the energy spectrum. When becomes larger than the charging energy
> E
C
, the charge transfer occurs only within the states with even number of
particles, and the 2e-periodicity characteristic for a superconducting system is
restored (Fig. 7.13(c)).
When also the leads become superconducting, the Josephson energy appears,
and the gaps open at the crossing points shown by black dots. The energy depen-
dence returns to Fig. 7.6 for a superconducting Josephson junction considered
before.
7.5. PARITY EFFECTS 143
Problem 7.1
Find the quasiclassical probability of MQT from the minimum of the wash-
board potential for I close to I
c
.
Problem 7.2
Find the lowest energy in the nearly free-phase limit, E
J
E
C
.
Problem 7.3
Derive Eq. (7.21).
Bibliography
[1] P.G. de Gennes Superconductivity of Metals and Alloys (W.A. Benjamin,
Inc, 1965).
[2] M. Tinkham, Introduction to superconductivity (McGraw-Hill, New York,
1996).
[3] J. Bardeen, L.N. Cooper, and J.R. Schrieer, Phys. Rev. 108, 1175 (1957).
[4] L.D. Landau and E.M. Lifshitz, Statistical Physics Part 2: Theory of the
condensed state, by E.M. Lifshitz and L.P. Pitaevskii (Pergamon, Oxford
1980), Chapter III.
[5] Abrikosov, A. A. Fundamentals of The Theory of Metals. (North Holland,
Amsterdam, 1998).
[6] D. Saint-James, G. Sarma, and E.J. Thomas Type II superconductivity
(Pergamon Press, 1969).
[7] R. Feynman, in: Progress in Low Temperature Physics vol. 1, edited by C.
J. Gorter (North Holland, Amsterdam, 1955) p. 36.
[8] A.A. Abrikosov, Zh. Eksp. Teor. Fiz. 32, 1442 (1957) [Sov. Phys. JETP 5,
1174 (1957)].
[9] P.W. Anderson, Journ. Phys. Chem. Solids 11, 26 (1959).
[10] A. F. Andreev, Zh. Eksp. Teor. Fiz. 49, 655 (1965) [Sov. Phys. JETP 22,
455 (1966)].
[11] G.E. Blonder, M. Tinkham, and T.M. Klapwijk, Phys. Rev. B, 25, 4515
(1982).
[12] S. Datta, Electronic Transport in Mesoscopic Systems, Cambridge Univer-
sity Press (Cambridge, 1995).
[13] C.W.J. Beenakker, Phys. Rev Lett. 67, 3836, (1991).
[14] V. Ambegaokar and A. Barato, Phys. Rev. Lett. 10, 486, (1963); 11,
104(E) (1963).
145
146 BIBLIOGRAPHY
[15] V. Ambegaokar and B. I. Halperin, Phys. Rev. Lett. 22, 1364, (1969).
[16] J.M. Ziman, Principles of the theory of solids, (Cambrgidge University
Press, 1965).
[17] G. Sch on and A.D. Zaikin, Phys Reports 198, 237 (1990).
[18] G.-L. Ingold and Yu. V. Nazarov, Charge tunneling rates in ultrasmall
junctions, in: Single Charge Tunneling, H. Grabert and M.H. Devoret,
eds., NATO ASI Series, Vol. 294, pp. 21107 (Plenum Press, New York,
1992); arXiv:cond-mat/0508728v1
[19] J.P. Pekola, K.P. Hirvi, J.P. Kauppinen, and M.A. Paalanen, Phys. Rev.
Lett. 94, 2903 (1994).

Aalto University Theory of Superconductivity

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aalto University Theory of Superconductivity

Uploaded by

Copyright:

Available Formats

Theory of Superconductivity

Tfy-3.4801 (6 cr.) P, III.

is the charge of the carrier. In superconductors the charge is carried

= 2e and the Cooper pair mass is 2m.

where is the mean free path.

1.4. CRITICAL CURRENTS AND MAGNETIC FIELDS 13

2 0.7) and type-II ( > 1/

The exact expression for H

1.5. QUANTIZED VORTICES 15

2. More favorable energetically

from below the Fermi surface, p

. Adding a particle with a momentum p and energy E above the Fermi

P = ih. The hole wave

p below the Fermi surface.

and has a wave function v

(r, )(r, )(r, ) (2.29)

operators which makes all calculations extremely complicated.

are then calculated using Eq. (2.31):

we nd that Eqs. (2.35) hold provided

(r)u(r) = v(r) (2.37)

(r, )(r, )(r, )

(r, )(r, )(r, )

(r, )(r, )(r, )

(r) (r, )(r, ))

u(r) = v(r) (2.59)

The condition of solvability of Eqs. (2.61) and (2.62) gives

Figure 2.7: The BCS spectrum of excitations in a superconductor. The solid

(q) states per spin and per unit volume

and of the group velocity v

in such a way that it vanishes

is called the condensation energy.

and subtract the complex conjugated

and subtract the complex conju-

The distribution function here is

Figure 3.1: The NS structure.

56 CHAPTER 3. ANDREEV REFLECTION

dp/2h = 2 since there is exactly one state

78CHAPTER 4. SUPERCONDUCTORINSULATORNORMAL-METAL INTERFACE

and in U, V with simultaneous change of 1 to

(N). The sum over the states is

(N). To get this distribution we note that the hole

(S). The sum over the states is

is given now by Eq. (4.45).

and the negative momentum but positive group velocity. They

(negative momentum and group velocity) and reected

= b. In other words, the

( = 1, 2) are the energies of the states in uncoupled superconductors

are normalized such that

the coecients obey

= 2e. The o-diagonal matrix elements describe transitions

Figure 5.4: A pendulum moves in time similar to variations of along the y

to jump over the barrier from the state

is now strongly suppressed by a considerably higher

6.1. ORTHODOX DESCRIPTION OF THE COULOMB BLOCKADE 117

is the characteristic charging energy of the island, and

dE [f(E) f(E +x)] = x

are the turning points satisfying E = U(). The probability of

couples the states at the Q = e boundary of the Brillouin zone

You might also like